Digital PDFs

EK-DNAPV-GD

September 1987

129 pages

Original

5.8MB

Document:	DECnet™ DIGITAL Network Architecture (Phase V) General Description
Order Number:	EK-DNAPV-GD
Revision:	0
Pages:	129
Original Filename:

OCR Text

DECnet™
DIGITAL Network Architecture
(Phase V)
General Description
Order No. EK-DNAPV-GD

September 1987

This document describes the design of the DIGITAL Network Architecture
that serves as a model for Phase V DECnet implementations. It includes
descriptions of functions. protocol messages. and operations.

To order additional copies of this document. contact your local
Digital Equipment Corporation Sales Office

digital equipment corporation. maynard. massachusetts

This material may be copied, in whole or in part, provided that the copyright notice below
is included in each copy along with an acknowledgment that the copy describes the Digital
Network Architecture developed by Digital Equipment Corporation.
This material may be changed without notice by Digital Equipment Corporation, and
Digital Equipment Corporation is not responsible for any errors which may appear herein.

The following are trademarks of Digital Equipment Corporation:
DDCMP
DEC
DECmail
DECnet
DECnet-DOS
DECnet-ULTRIX
DECnet-VAX
DECUS
DNA
LAN Bridge
MAIL-II

Message Router
PDP
ReGIS
RSTS/E
RSX-ll
RSX-IIM-PLUS
ThinWire
ULTRIX
UNIBUS
VAX
VAX CDD

VAX SPM
VAX VALU
VAX VTX
VAXcluster
VMS
VT100
VT200

~D~DamD

IBM is a registered trademark of International Business Machines Corporation.
MVS is a trademark of International Business Machines Corporation.
MS, MS-DOS and Microsoft are registered trademarks of Microsoft Corporation.

Contents
vii

Preface
1 Introduction
1.1
1.2
1.3

1.4
1.5

1
2

Necessity for an Architecture
Design Goals . . . . . . . . .
The Structure of DNA . . ..
1.3.1 Layered Architecture .
1.3.2 The Layers of DNA
1.3.3 Message Flow in DNA
1.3.4 The Naming Service
1.3.5 Network Management
Relationship of DNA to OSI .
Relationship to Phase IV

2
4
4
6

9
10
10
11
13

2 Physical Layer
2.1
2.2
2.3
2.4

Physical Layer Functions . . . . . . . . .
Physical Layer Functional Modules .. .
Modem Connect Functional Description
2.3.1 DNA Modem Connect Features.
X.21 Functional Description
2.4.1 DNA X.21 Features

3 Data Link Layer
3.1
3.2
3.3
3.4

21
21

Data Link Layer Functions
Data Link Layer Protocol Modules
DDCMP Functional Description
3.3.1 DDCMP Features . . .
HDLC Functional Description.
3.4.1 HDLC Features.

22
23
23

26
26

4 Local Area Networks
4.1
4.2
4.3

16
16
17
17
18
18

Services Provided . . . . . . . . . . . . .
Characteristics of the CSMA/CD LAN.
Relationship to International Standards
III

30
30

4.3.1 Addressing .. . . . . . . . . . ..
4.3.2 MultIplexing on ISO 8802-3 LANs
Ethernet . . . . . . . .
Bridges
4.5.1 Bridge Services
4.5.2 Extended LAN Topology
4.5.3 Bridge Algorithms . . . .
4.5.4 Bridge Management . . .
ISO 8802-3 and Ethernet Coexistence.

30
31
33
33
34
34
34
35
35

5 Network Layer
5.1 Network Service Characteristics.
5.2 Network Layer Concepts ..
5.2.1 Hierarchical Routing
5.2.2 System Types .. .
5.2.3 Addressing . . . . .
5.2.4 Subnetwork Types .
5.2.5 Multiple Routing Domains - Static Routing
5.3 Network Layer Functions . . . . . . . . . .
5.3.1 Subnetwork Independent Functions.
5.3.2 Subnetwork Dependent Functions.
5.4 Routing Operation . . . . . .
5.4.1 The Decision Process ..
5.4.2 The Update Process ..
5.4.3 The Forwarding Process

31
38
38
38

Transport Layer
6.1 Transport Layer Functions.
6.2 Transport Layer Protocols .
6.3 OSI Transport Protocol . .
6.3.1 Connection Establishment
6.3.2 Data Transfer.
6.3.3 Disconnection.
6.4 NSP
6.4.1 Connection Establishment
6.4.2 Data Transfer.
6.4.3 Disconnection.

Session Control
7.1 Functional Description . . . . . . . . . .
7.2 Session Control Functional Components
7.3 Connection Control . . . . . . . . . . . .
7.3.1 Requesting a Connection by Destination Address
7.3.2 Receiving a Connect Request

55
55
55
56

4.4
4.5

4.6

39
39

41
42
43
43
44
44
44
45
45

47
48

49
50
50
52
52
52
52
53

57
57

7.4

7.5

7.3.3 Sending and Receiving Data. . . . . . . . . . . . . .
7.3.4 Monitoring a Transport Connection . . . . . . . . .
7.3.5 Disconnecting or Aborting a Transport Connection.
Address Resolution . . . . . . . . . . . . . . . . . . . . . . .
7.4.1 Towers . . . . . . . . . . . . . . . . . . . . . . . . . .
7.4.2 Establishing Protocol Sequences for Communication
7.4.3 Maintaining the Towers in the Namespace . . . . . .
Address Selection . . . . . . . . . . . . . . . . . . . . . . . .
7.5.1 Requesting a Transport Connection by Destination Name

58
58
58
59
59
60
61
61
61

8 Naming Service
8 .1 Naming Service Concepts . . . . . .
8.2 The Semantics and Syntax of Names
8.2.1 The Semantics of Names.
8.2.2 The Syntax of Names .
8.3 Contents of Namespace Entries .
8.3.1 Global Attributes . . . . .
8.3.2 Predefined Object Classes
8.4 Operational Concepts and Terminology
8.5 Functional Decomposition of the Naming Service
8.5.1 Clerks . . . .
8.5.2 Nameservers

63
63

73
73

Support of X.25
9.1 X.25 Service Functions .
9.2 X.25 Modules . . . . . .
9.3 X.25 Packet Level . , .
9.3.1 Basic Features of the Packet Level
9.3.2 Optional Facilities of the Packet Level
9.4 X.25 Gateway Access . . . . . . . . . . .
9.4.1 X.25 Gateway Access Operation

64
64
64
65
66
66
66
69
69
71

74
74

76
77

10 Network Management
10.1 DNA Network Management Model
10.1.1 Entity Hierarchy and Naming
10.2 Network Management Operation . .
10.2.1 Common Management Information Protocol (CMIP) .
10.2.2 Event Logging . . . . . . . . . . . . . . .
10.2.3 Maintenance Operations Protocol (MOP)
10.2.4 Network Control Language (NCL)

81
82
83

11 DNA Applications
1l.1 Heterogeneous File Access and Transfer
11.2 Network Virtual Terminals
1l.3 Electronic Mail . . . . . . . . . . . . . .

84
84

85
85
86

87
88
89

11. 3 .1 Mail-II . . . . . . . . .
11.3.2 Message Router. . . . .
11.4 SI\ A Interconnect Applications
11.4.1 SI\A Gateway Access .
11.4.2 3270 Terminal Emulator
11.4.3 Remote Job Entry ..
11.4.4 Data Transfer Facility
11.5 VMS Services for MS-DOS
11.6 Time Service . . . . . . . . .
11.7 Computer Conferencing .. .
11.8 VAX System Performance Monitor
11.9 Videotex . . . . . . . . . . . .
11.10Distributed Queueing
11.11Remote System Management

89
89
89
90
92
92
92
93
93
93

94
94
95
95

12 0 SI Upper Layers and Applications
12.1 General OSI Upper Layer Components.
12.2 OSI Applications . . . , . . . . . . . . .

Glossary

101

Index

114
114

Index

97
98

Preface
This book is an overview of the DIGITAL Network Architecture (DNA). DNA is a model
of structure and functions upon which DECnet implementations are based. DECnet is a
family of communications software and hardware products that enable DIG ITAL operating
systems and computers to function in a network with other DIGITAL systems and with
systems manufactured by other vendors.
A DECnet network is a group of computer systems with associated operating systems,
DECnet software, and communication hardware that are connected to each other by physical channels, or lines. Each computer in a network containing a DECnet implementation
is called a system. A network therefore consists of connected systems and lines. DNA
defines standard protocols, interfaces, and functions that enable DECnet systems to share
data and access others' resources, programs, and functions.
The DNA Model incorporates both a set of proprietary protocols and interfaces, and a
set of protocols and services defined for Open System Interconnection by the International
Standards Organization.
This book describes Phase V of DNA. Phase V provides compatibility with the previous
version of DNA, Phase IV.
DNA supports a broad range of applications and a variety of network topologies. (A
network topology is a particular configuration of systems and lines.) User documentation and marketing brochures describe in detail various types of application and specific
programming and network management information.
This document summarizes the design and structure of DNA and serves as an introduction to the DNA functional specifications. It is intended for readers with a knowledge
of communications technology who desire an understanding of the overall DNA structure.
A glossary at the end of the document defines many DNA terms. Additionally, many DNA
terms are italicized and explained in the text at their first occurrence.
Phase V of DNA defines a rich set of functions. Some implementations will not need
all of these functions. There are also some aspects of Phase V which may not be present
in initial implementations.

Vll

Vlll

Chapter 1

Introduction
The Digital Network Architecture (DNA) defines the means by which computer systems
can communicate with each other to form a distributed processing system. It specifies:
• the communication protocols by which systems exchange information, including the
rules for constructing and interpreting messages.
• the internal interfaces which allow one protocol to be described and built in terms of
the services provided by another.
• the policies and mechanisms by which systems adapt to changing loads and configurations to make the best use of available resources.
• the means by which a distributed network can be managed, both locally and remotely.
This General Description relates to Phase V of the DNA model. Phase V is an evolutionary
step from the previous version, Phase IV. There are two major new technical directions in
Phase V: integration of the OSI standards, and support for increasing network sizes.
OSI ("Open Systems Interconnection") is a set of international standards developed
by the International Standards Organization (ISO). Its goal is to allow different vendors'
computer systems to be freely interconnected. The standards for the lower layers, providing
the basic communications functions, are now substantially complete. DNA uses these OSI
standards wherever possible so that Digital computer systems can be used in multi-vendor
networks and distributed processing systems.
The increasing importance of networking means that computer networks are continually
getting larger. Where a few years ago networks of even a dozen systems were rare, today
networks of hundreds of systems are commonplace and several much larger networks, up to
tens of thousands of systems, are in use. Furthermore, this increase in the size and number
of networks means that they can no longer be operated exclusively by specialists as they
have in the past. DNA includes many capabilities to simplify the design, construction
and operation of large networks. Of particular importance is the Naming Service, which
automates the distribution of critical shared information such as the network addresses of
resources, and their protocol capabilities.
1

CHAPTER 1. INTRODUCTION

1.1

Necessity for an Architecture

A system as complex as a DECnet network could not be built without an architecture.
The architecture is necessary for several reasons:
• There is a wide variety of computer hardware, communication devices and operating
systems. These are built by separate implementation groups. Communication is only
possible if they use a common set of specifications.
• Industry Standards (in particular, OSI) are a vital part of computer networking.
The architecture specifies how these standards are used and permits individual standards to be introduced and integrated without disrupting the overall function of the
network.
• Communications technology is constantly developing. A network product family
needs to be able to evolve to include these new technologies. The architecture makes
this possible by imposing a modular structure whereby changes in one area do not
affect other areas.
• The modular structure of the architecture permits high-performance functions to be
implemented in hardware, in a system-independent way. For example, Local Area
Network adapters implement the LAN data link protocols.

1.2

Design Goals

DNA was designed to support the following goals.

Conceal network operation from the user. The operation of a large network is necessarily complex, but to the user it should appear simple. It then becomes very
natural to build distributed applications.
Support a wide range of applications. Almost any application may be distributed
across a network. DNA supports a great variety of applications, and allows new
ones to be developed freely.
Support a wide range of communication facilities. The technology of data communications is constantly evolving. Not only are greater speeds becoming available, but
the nature of the services is also changing. This is clearly illustrated by the development of Local Area Networks and X.25 packet-switched networks. The structure
of DNA allows these new technologies to be integrated and supported without major
change to the remainder of the architecture. In Phase V, DNA supports traditional
synchronous data links, Local Area Networks, X.25 packet-switched networks and
X.21 circuit-switched networks.
Support a wide range of network topologies. DNA does not impose any particular
topology (such as star, mesh or hierarchical). It makes the best use of a topology
constructed to meet the performance and cost requirements of each network.

1.2. DESIGN GOALS

Make maximum use of standards. Standards are critical to successful computer networks. Wherever feasible, DNA uses OS1 standards rather than proprietary means to
attain its goals. Where OS1 standards are being worked on but are not yet approved,
DNA is designed to make future migration as straightforward as possible.
Require minimum management intervention. Wherever possible, DNA allows a network to operate without operator ot network management intervention. This is particularly important for user-oriented systems, such as personal workstations. The
architecture includes algorithms and mechanisms for automatic determination of operational characteristics. Examples include the dynamic routing algorithm, and the
automatic determination of a system's address.
Be manageable. Some functions cannot be automated, and for those which can it is
sometimes necessary to override the automatic choice. This is particularly true for
large or complex networks. DNA permits remote management of all network functions. The architecture does not impose any particular style of management (such
as centralized or hierarchical), but leaves this to the user's choice.
Permit growth. DNA allows a network to grow from just two systems to a huge global
network without disruption or major reconfiguration. To the systems attached to the
network, such growth is completely transparent.
Permit migration. As new versions of the architecture and its implementations are developed, existing networks must migrate to them. Migration of a large network may
take months or even years, and during this period it must continue to operate normally. Each Phase of DNA will fully interwork with the previous Phase. All Phase V
systems can communicate with the Phase IV systems in a network throughout its
migration period.
Be subsettable. Not all systems need to implement all functions of the architecture.
DNA allows an implementation to omit certain functions, particularly those which
are expensive to provide and which do not need to be in every system for the network
to function.
Be extensible. DNA allows for the incorporation of future technology changes in hardware, software and standards.
Be highly available. DECnet networks can be configured to maintain operation even if
some lines or systems fail, with minimum prior planning or operator intervention.
Be highly distributed. The major functions of DNA, such as routing and network management, are not centralized in a single system in the network. This in turn increases
the availability of the network.
Allow for security. DNA allows for security at several levels. User access control IS
implemented in most DECnet products.

CHAPTER 1. INTRODUCTION

1.3

The Structure of DNA

1.3.1

Layered Architecture

DNA is a layered architecture. This means that the functions required to achieve its goals
are divided into related and logically coherent groups called layers. The layers are built
on top of one another, so that one layer makes use of services provided by the one beneath
it. The detailed protocols and other mechanisms of a layer are hidden from the layers
above and below - all that is seen by the layer above is the service it provides. This
gives the property of layer independence, so that very substantial changes can be made to
one layer (such as changing the protocol used) without affecting other layers. A layered
architecture is much better able to meet the goals set for DNA than one in which functions
are performed in a less structured manner.

I
---t--t
- - - - - ---+-tI n-layer client

n-serVlce

n-protocol
n-entity

n-entity

------+--- -

----~-----

n-layer

(n-l )-service

Communication through lower layers

Figure 1.1: Principles of a single layer
Figure 1.1 illustrates the principles of a layer, using the terminology of the OSI Reference
Model. It applies to any layer - for generality, the term n-layer is used to refer to an
unspecified layer of the architecture. Part of the layer exists in each of the communicating
systems. Within one system, the component responsible for the n-layer is called the nentity. The two n-entities communicate with each other using a protocol, called the nprotocol. This protocol is conveyed using the services of the next lower layer. All layers
provide a data transfer service, although the details vary. Some layers provide additional
services such as connection management. In general, the n-protocol is carried as data by
the (n-l)-service. The lower layer has no knowledge of, or interest in, the content of the
data it conveys. Each n-entity provides the n-service to a local client. The client may be
either the (n+l)-layer, or an end-user.
When a layer carries data on behalf of a client, it also adds a header containing Protocol
Control Information (PCI). The PCI contains information which the n-entities need to
control each others' operation. It typically includes message sequence numbers and information about the precise service requested. The combination of the PCI and the user data

1.3. THE STRUCTURE OF DNA

(n-l )-layer

n-layer

(n+ 1)-layer

PCI

{n+l)-layer
User Data
n-layer User Data
nPDU

Figure 1.2: Nesting of Protocol Control Information

is called a Protocol Data Unit (PDU) - a PDU of layer n is called an nPDU. Thus a PDU
of the Transport layer, for example, is called a TPDU. Often, other names are used for
PDUs at a specific layer, such as packet or frame. Since there are several layers involved
in providing a service to the user, the data actually sent across a physical channel includes
the PCI for several layers, each carried as data by the lower layer. This is illustrated in
figure 1.2. In the Data Link layer, some PCI is also carried in a trailer which follows the
user data.
The data passed to a layer by its client is called a Service Data Unit (SDU). SDUs
pass within a system, between a layer and its client. They are not visible from outside the
system. In contrast, PDUs pass between systems, between the communicating entities of a
layer. They are visible outside the system as the data flowing "across the wire" between the
two communicating systems. Section 1.3.3 describes this in terms of the layers of DNA. In
the simplest case, a PDU consists simply of the corresponding SDU together with a header
containing PCI. Some layers have the ability to divide an SDU into smaller units, sending
each in its own PDU. For example, this allows a layer to convey very large SDUs even
though the layer beneath it has a relatively small maximum SDU size.
In DNA, a layer may use more than one protocol, for reasons of compatibility with earlier versions or to accommodate alternative option selections from standards. In addition,
in some layers one protocol may be a client of another within the layer. A module is an
entity which implements a specific protocol, whether or not it is the only protocol in its
layer. Each protocol in DNA is defined as a module. A module provides a service to its
clients, even if they may be further modules within the same layer.
Layers and modules generally have parameters which can be selected by a client, for
example at the time a connection is made. Often, a client needs a particular service but
does not need to be concerned with the details of individual parameters. DNA provides
templates in many modules, which are named collections of parameters and their values
defined using network management. A client can reference a template by its name, gaining
the same effect as if it had specified the parameter values individually. This gives a further
degree of layer independence, since the architecture for one module need not reflect in
detail the parameters of a supporting module. For example, X.25 provides many options
to do with the exact way the X.25 connection is made. Users of X.25, such as the Network

CHAPTER 1. INTRODUCTION

Routing mod ule, can access these without having to reflect the options explicitly in their
own definition.

1.3.2

The Layers of DNA

OSI Application
DNA Application
Application
Presentation
Naming
Service

DNA
Session Control

Session

Transport
Network
Data Link
Physical

Figure 1.3: The Layers of DNA
Figure 1.3 shows the layers of DNA and their relationship. These are further described
below.

Physical layer. The Physical layer is concerned with transmission and reception of data
on the transmission medium. Its functions include:
• Conversion of data to and from electrical signals.
• Monitoring signals from the communication channel, such as modem signals.
• Bit level synchronization, either by direct generation and recovery of the clock
signals or using an external clock (for example, from a modem).
• Management of Physical layer connections, such as dialup lines and X.21 circuits.
The Physical layer provides the mechanisms for this but the policies (when to
initiate and terminate a circuit) are administered by the Network layer.
The service provided by the Physical layer to its client is concerned with the transmission and reception of individual bits that make up higher-layer messages. In
this layer, detailed operation is fully described by standards such as EIA RS-232-D,
CCITT V.24 and X.21, and ISO 8802. DNA specifications are therefore concerned
only with network management and with interfaces to other components of the architecture.

1.3. THE STRUCTURE OF DNA

Data Link layer. The Data Link layer provides a dependable communication path between directly-connected systems in a network. It operates protocols which detect
errors induced in the Physical layer (for example by electrical disturbances). This
layer also corrects such errors where the anticipated error-rate is high, for example
on synchronous lines.
DNA defines three protocols for this layer. DDCMpTM and HDLC operate over
synchronous lines. DDCMP is a Digital-defined protocol used in previous Phases of
DNA, which can also operate over asynchronous lines. HDLC uses selected features
of the ISO standards. A subset of this protocol is compatible with LAPB, used in
X.25. Over Local Area Networks, DNA uses ISO 8802-2 and -3, which are identical
with the corresponding parts of IEEE 802, and can also use the Ethernet standards
for compatibility with Phase IV.

Network layer. This layer routes user data between systems in the network. In DNA,
this layer provides the Connectionless-mode Network Service (CLNS) of as!. Each
piece of user data (that is, each NSDU) travels through the network in a self-contained
packet, containing its destination and source addresses and other required information. The protocol used is the ISO Internetwork Protocol (ISO 8473). This layer
exists both in the communicating systems and also in the routers which join together
the data links forming the network. The Network layer also maintains the databases
needed to find routes through the network from a source to a destination. To make
this possible, it operates a dynamic adaptive routing algorithm which automatically
adapts the routes to the current topology of the network. This algorithm is defined
in the DNA Network Routing specification. It allows for networks of several hundred
thousand systems. The Network Routing specification also contains the mechanisms
for using the various kinds of communications media, such as LANs, synchronous
circuits and X.25 networks, and for interworking with other (non-DNA) networks
that also use the OSI protocols.
DNA also provides the Connection-mode Network Service (CONS) in systems which
are directly attached to an X.25 Wide Area Network, or which have access to such a
network through the X.25 Gateway Access Protocol.

Transport layer. This layer provides a reliable end-to-end service between communicating systems, concealing from its users the detailed way in which this is achieved.
This layer (and all higher layers) operates only in the communicating systems. Its
functions include:
• Recovery from data loss, where the Network layer fails to deliver a packet (for
example due to congestion), by retransmission.
• Flow control, to match the transmission rate to the reception rate.
• Congestion avoidance - using information provided by the Network layer, the
Transport layer manages its use of the network to avoid saturation of the network
while providing a fair service to all users.

CHAPTER 1. INTRODCCTION

• Segmentation and reassembly of user messages, so that there is no limit to
the size of user messages (TSDUs) even though the size of network messages
(NSDUs) is limited.
DNA contains two protocols for this layer. NSP (Network Services Protocol) is a
Digital-defined protocol which was first defined for DNA Phase I. The 051 Transport
Protocol (ISO 8073) is used for the first time in Phase V. Digital systems use Class 4 of
this protocol, which can operate over both the Connectionless-mode and Connectionmode Network Services. DNA also describes the uses of Classes 0 and 2, for use only
over the CONS.
Above the Transport layer, DNA permits two alternative modes of access. Proprietary and
user-developed application protocols use the DNA Session Control layer. Standardized
application protocols use the OSI Session, Presentation and Application protocols.

DNA Session Control. This provides the logical links (that is, connections) which are
seen by the application programs. For data transfer, these provide the functions of
the Transport service. Session Control's additional functions are concerned with the
initiation of a logical link and its management within a system. The most important
functions of this layer whose operation is specified by the architecture are:
Name to address translation. Remote objects are addressed by their object name.

Session Control translates the name into the addresses used at each of the protocol layers, and in particular the Network address which allows the Network layer
to send traffic to the correct remote system. For compatibility with Phase IV,
objects may also be addressed using a node name which identifies the remote
system, and a means to identify an object within the system.
Protocol selection. At several layers, DNA provides a choice of protocols. Partly, this

is for compatibility between Phases, and partly it is to permit communication
with non-DNA systems which have different choices in their selection of OSI
options. When Session Control performs the object name translation, it also
obtains information about the protocols which can be used. This is compared
with the capabilities of the local system, and a mutually compatible protocol
set is selected.
Access control. Session control provides features whereby remote users can be identi-

fied. In conjunction with system-specific access control mechanisms, this allows
access to particular objects to be restricted to a limited set of remote users.
In addition, the Session Control layer is where system-specific logical link initialization functions are performed. These include process addressing and process activation.

OSI Upper Layer Protocols. The OSI Session, Presentation and Application layers are
used to support standardized application protocols. DNA does not specify the operation of these layers, since they are fully described in the corresponding standards.

1.3. THE STRUCTURE OF DNA

User access IS In general provided at the application layer, although some implementations may provide direct access to the services of the Session and Presen tation
layers.
User access to DNA modules is not restricted to the DNA and aS! application layers. For
example, there are several applications which make direct use of the Data Link layer, to
meet specialized requirements in a LAN environment.

1.3.3

Message Flow in DNA

TSDUs
TPDUs

Transport
Layer

-------~--------------------------------------------------------

NSDUs

NPDUs

(packets)

Router
...-----------,
to4-t-------+t

Network
Layer

-------~--------------------------

DPDUs

(frames)

-------

Data Link
Layer

----------------------------- -----------------------J
I
Physical channel

Physical channel

Physical
Layer

Figure 1.4: DNA Message Flow
Figure 1.4 shows how messages flow between DNA systems. At the Data Link layer, communication is strictly between directly connected systems. The messages passing between
entities at this layer are Data Link PDUs (DPDUs), which are often called frames. At
the Network layer, routers pass traffic from one data link to another on the path from
one system to another. The messages at this layer are Network PDUs (NPDUs), often
called packets. Routers examine the NPDUs to determine the next part of the route, and
update some parts but largely pass them on unchanged. NPDUs passing between Network
Layer Entities correspond to Network Service Data Units (NSDUs) which pass between
the Network layer and the Transport layer.
At the Transport layer, only the end systems are involved. The messages passing
between the Transport entities are Transport PDUs (TPDUs). They are not examined by
the routers, but pass through transparently as user data. Data passing between Transport

CHAPTER 1. INTRODUCTION

and its users is in Transport Service Data Units (TSDUs). The user will normally be either
DNA Session Control or the OSI Session layer.

1.3.4

The Naming Service

The name translation and protocol selection functions of the DNA Session Control layer
require a large database containing information about all systems in the network. It would
be an impractical task for each system to keep its own copy of this database, particularly in
a large network. DNA therefore includes a network-wide Naming Service, which can store
all names having network-wide significance. Any name can be accessed and its associated
attributes (such as network addresses) can be retrieved, from any point in the network.
Every name is unique. A hierarchical directory structure (similar to a filestore directory
structure) simplifies the task of assigning unique names.
N ames are physically stored in directories. In general, a directory is replicated, that
is, stored on more than one physical system. This protects against hardware failure (for
example, disk crashes) and also makes the information accessible even during temporary
network problems. The Naming Service contains the mechanisms needed to ensure that
the copies of directories become synchronized as changes occur.
A major use of the Naming Service is to maintain the database of object names, and
associated addresses and protocol selection information, which Session Control uses. However, it is not constrained to this application, and can be used by any application which
requires names to have unique network-wide significance.
The Naming Service is both a component of DNA and also a user of it. Communication
between the Name Servers which provide the service uses the normal services of DNA
through Session Control.

1.3.5

Network Management

DNA provides powerful distributed network management facilities which allow network
components to be managed remotely. The network management facilities available include:
• Configuring the network components within a system, for example to determine how
its physical links are to be used.
• Setting operational parameters such as timer values and resource limits.
• Examining configuration information and the operational parameters.
• Examining counters of diagnostic information such as messages sent and received and
error conditions.
• Generation of events which can be collected and analyzed at a remote site, such as
circuit failures and recovery.
• Diagnostic operations, in particular loop back of data to allow a failing component to
be isolated.

1.4. RELATIONSHIP OF DNA TO OSI

Each system in the network, and each component (such as a protocol module) within a
system, contains a network management agent. The agent accesses information held by
the component, to report on it or to modify it. The information which is kept by each
component, and how it may be modified and used, is specified by the corresponding part
of DNA. Each system has a unique agent called the node agent which receives requests
from other systems and dispatches them to the appropriate part of the system.
Network management is used through a director. DNA defines a network management
command language called NCL which corresponds closely to the architected structure of
a system. A program which provides an NCL interface to a user is an example of a
director. A director may either be local (that is, running on the system it is managing) or
remote. A local director can be used to perform management operations when the system
is not yet part of a network, for example during system initialization. A remote director
communicates with the system it is managing using the following protocols.

Conunon Management Information Protocol (CMIP). This is used for communication betwee~ a director and the target node agent. It provides an encoding of
the network management operations which can be performed, and their parameters.
For example, an operation might be "read counter" and the parameters would be
the name of the counter and the name of the system component in which it resides.
CMIP is a client of the Session Control layer , and operates in the same way as any
other DNA application protocol.
Maintenance Operations Protocol (MOP). This is used for low-level communication
when a system is not fully operational. It is a client of the Data Link layer. Its most
important functions are the triggering and execution of down-line load and up-line
dump, and link-Ievelloopback functions.
Event Logger Protocol. This is a particular fonn of CMIP, used by the node agent to
report events when this has been requested by a director. Like CMIP it operates as
a DNA application protocol.
Network Information and Control Exchange (NICE). This is the protocol used by
Phase IV for management communication. It is supported by NCL for management
of Phase IV systems during the migration period.

1.4

Relationship of DNA to OSI

As Figure 1.3 shows, the layers of DNA are in direct correspondence with the seven layers
of the OSI Reference Model (ISO 7498). This reflects both the origin of OSI, which derived
from contemporary network architectures including DNA, and the continuing development
of DNA to align it more closely with international standards. In Phase V, the protocols
used up to the Transport layer are the OSI protocols, earlier proprietary protocols being
retained as an alternative for compatibility purposes. Specifically, the following major OSI
standards are used:

CHAPTER 1. IXTRODUCTION

Physical layer. The DNA Physical layer has always been based on available standards,
which are implemented in hardware devices. These include, EIA RS-232-D (and
the corresponding ISO standards and CCITT Recommendations) and the Ethernet
standards as now reflected in ISO 8802-3 and IEEE 802.3.
Data Link layer. The version ofRDLC which is used in DNA uses ISO 4335 and ISO 7809.
For Local Area Networks, DNA uses the Logical Link Control protocol of ISO 8802-2
and IEEE 802.2.
Network layer. The ISO Protocol for Providing the Connectionless-mode Network Service, ISO 8473, is used for data transfer. Exchange of routing information between
end-systems and routers uses the ISO End System to Intermediate System Routing
Protocol, ISO 9542. The Connection-mode Network Service, where used, is provided
using the X.25 Packet Layer Protocol, as defined in ISO 8208 and the mapping defined
in ISO 8878. The addresses used by the DNA Network layer follow the corresponding
OSI standard, ISO 8348 Addendum 2.
Transport layer. DNA includes Class 4 of the OSI Transport Protocol, ISO 8073. Classes
o and 2 are also available, for use over the CONS. The service provided to a DNA
Application corresponds to the OSI Transport Service, ISO 8072. This service is also
used by applications which use the OSI upper layer protocols.
Above the Transport layer, OSI protocols are available as an alternative but are not used
for all applications. There are several reasons for this:

Functions available. DNA supports many applications and their protocols. Only a few of
these are under study for OSI standardization.
State of the standards. At the time of writing, the only OSI upper layer standards which
have been approved as an International Standard are those for the Session layer. The
others are at various stages of drafting, and are therefore subject to further change
before final approval.
Compatibility. Communication with previous implementations is essential and can only be
done using the proprietary protocols.
In Phase V of DNA, specific implementations of upper layer standards will be included as
they become available. Development of an integrated DNA architecture for the OSI upper
layers will take place during the lifetime of Phase V, reflecting the progress of the standards
themselves.
The DNA network management protocols are based closely on early drafts of ISO's
work on the subject. The CMIP protocol used in DNA is based on drafts of the ISO CMIP
protocol (ISO DP 9596), and uses the OSI Remote Operations Service (ISO DP 9072) and
the OSI Presentation protocol (ISO 8823 and ISO 8825).
The functions of the DNA Session Control layer are not yet specified for OS1. When
they are, it is expected that they will reside in the OSI Application layer. Integration
of these functions in the DNA use of the OSI Application layer will form part of future
extensions to DNA.

1.5. RELATIONSHIP TO PHASE IV

1.5

Relationship to Phase IV

In keeping with the goals of DNA, a system which implements Phase V is fully able to
communicate with a Phase IV system. A Phase IV network can be migrated gradually
to Phase V without requiring closely synchronized changes to multiple systems. All of the
protocols used in Phase IV are included in Phase V systems, although in many cases they
will be used only for communication with Phase IV systems. To the user of a DECnet
network, the change from Phase IV to Phase V will not be apparent, except that the new
functions of Phase V will gradually become available.
A summary of the major changes between Phase IV and Phase V is given below.

Terminology. The names of some layers have changed from Phase IV:
• The Network layer was called the Routing layer.
• The Transport layer was called the End Communications layer.
• The Physical layer was called the Physical Link layer.
These changes have been made for alignment with the names used in as!.

Local Area Networks. Phase V uses the frame format defined by ISO 8802-2 and -3 and
IEEE 802.2. This differs slightly from the Ethernet frame format used in Phase IV.
DNA systems automatically use the correct format for communication with other
systems on the same LAN. A Phase V system always transmits and receives in ISO
format, and also listens for frames in Ethernet format. If there are Phase IV systems
on the LAN (that is, if Ethernet format frames are received) then it also transmits
using Phase IV format when necessary.
Data Link protocols. Phase V supports HDLC as well as DDCMP. A circuit must be
configured to use the same protocol as the remote system, otherwise it will fail to
initialize.
Network layer. There are many changes to the Network layer in Phase V, to accommodate the OSI standards and to permit larger networks. These are described in
Chapter 5.
Transport layer. Phase V supports the OSI Transport Protocol as well as NSP. A common interface is presented to the user of the Transport layer. Selection of the correct
protocol, for communication with another DNA system, is performed by the Session Control layer. For communication with non-DNA systems, the OSI Transport
Protocol is always used.
Session Control layer. The major change to this layer is the use of the Naming Service
for translation of object names and obtaining protocol selection information. Application programs which use this layer (which is the layer normally exposed in DECnet
products) need not change in Phase V, but some new facilities available through the
Naming Service will only be available if corresponding new facilities of the interface
are used.

CHAPTER 1. INTRODUCTION

Network Management. Both the protocol and the command language (NCL) are new
for Phase V. The Phase IV command language (NCP) and protocol (NICE) are
available for communication with Phase IV systems.

Chapter 2

Physical Layer
The Physical layer provides mechanical, electrical, functional and procedural means to activate, maintain and deactivate physical connections between directly connected systems,
and for Physical layer clients to transfer data. Functions of the layer include encoding and
decoding signals on the physical interface, bit synchronization of received data, bit transmission, and interfacing the communications channel to the processor and memory used to
implement higher layer protocol functions. Implementations of this layer encompass hardware interface devices and device drivers in operating systems, as well as communications
hardware such as modems, transceivers, and the physical communications links themselves.
Protocols for the Physical layer are rudimentary. No special Physical layer standards
have been developed for DNA. Instead, it relies on industry standards for the Physical
layer, thereby ensuring that DECnet products can operate over available technologies and
physical networks. Physical layer standards supported by DNA for wide area networks
include the EIA RS-232-D and RS-423 specifications, and the CCITT V.24 and X.21 bis
specifications. Physical layer standards supported by DNA for Local Area Networks include
two baseband implementations of the ISO 8802-3 LAN, the original and the Thin Wire™
specifications, and a broadband implementation.
Phase V of DNA defines the functions and operation of the Physical layer more completely than earlier versions of the architecture. The way in which industry standards for
the layer are integrated into DNA is fully described. Network management of the Physical
layer, which in Phase IV was bound to the management of Data Link layer protocols, has
been rationalized and enhanced.
DNA support for CCITT Recommendation X.21 has been added to Phase V, allowing
implementations of DECnet access to and use of the facilities of X.21 circuit switched
public data networks. The architecture for X.21 has been closely aligned with that for
modem connect type lines, such that a common set of services is presented to the higher
layers of DNA.
The DNA Physical layer is structured in a modular way, allowing support for other
Physical layer standards, such as those for ISDN, to be added in later versions.
15

CHAPTER 2.

2.1

PHYSICAL LA YER

Physical Layer Functions

The following functions are performed by Physical layer modules.

Connection establishnlent and release. These functions are provided where a communications link is attached to a circuit switched wide area network such as an X.21
or public telephone network. They allow physical connections to be dynamically
established and released.
Bit synchronization. This function establishes synchronization with an incoming bit
stream, and thereafter clocks data in from the communications channel at the correct
rate.

Data transfer. The Physical layer provides a bit stream service interface where data is
passed across the interface in I-bit units. On the transmit side, data bits passed by
the client are clocked onto the communications channel. Conversely on the receive
side, bits are clocked in from the channel and presented to the client. The Physical
layer maintains sequence; data bits are delivered in the same order in which they were
submitted. A communications link may allow full-duplex or half-duplex transmission.
Fault notification. This function reports fault conditions to Physical layer clients.
Management. These functions allow network management to control and monitor the
operation of the Physical layer, for example to set operating characteristics of communication links, activate and deactivate physical connections, monitor the status of
active connections and perform loopback. Physical layer modules maintain management counters and report significant events.

2.2

Physical Layer Functional Modules

The DNA Physical layer contains a number of functional modules, one for each type of
communications interface supported by DNA. These modules interface with the communications channel and make Physical layer services available to the higher layers of DNA.
Many of the functions performed at the Physical layer are common across all the modules
although details vary in each case. The modules comprise functions to get and set DTEDCE interchange circuit signals, where applicable, and functions to transmit and receive
data. At this level, internal system requests are mapped onto electrical signals on the
communications channel. These functions are media specific. For example, the transmit
function for a CSMA/CD LAN differs from that for an X.21 interface.
Currently, the DNA Physical layer defines functional modules for the following communications interfaces:

• Modem Connect. This is described in Section 2.3.
• X.21. This is described in Section 2.4.
• CSMA/CD LAN. This is described in Chapter 4.

2.3. MODEM CONNECT FUNCTIONAL DESCRIPTION

2.3

Modem Connect Functional Description

DNA Modem Connect is the name given to the Physical layer module which defines how
DNA operates over communication links conforming to industry standards for modem
connection. There are a number of these standards, each with different characteristics.
However, the standards encompassed by DNA Modem Connect are sufficiently similar to
be covered by a single architectural module. Differences between the standards are treated
as variants within this general module for modem connection.
DNA Modem Connect defines how the Physical layer services provided by switched
and leased lines are integrated into DNA. It defines the mechanisms which allow higher
layers of DNA to use these services, and for network management control and monitoring
of lines. The module supports a number of the industry standards for the layer, including
EIA RS-232-D and CCITT V.24, and CCITT V.25 and V.25bis for auto call and auto
answer. The module also defines null modem operation, where only the data interchange
circuits are used.

2.3.1

DNA Modem Connect Features

This section describes some of the major features of DNA Modem Connect.

2.3.1.1

Call Control Services

DNA Modem Connect defines services which permit the control and monitoring of switched
line connection establishment and release. This allows clients to make outgoing calls, handle
incoming calls and clear calls. The progress of an outgoing call attempt may be monitored,
and reason information is returned if the call fails to connect. Call control services do not
apply to leased line connections.

2.3.1.2

Data Transfer Services

DNA Modem Connect defines services which permit the higher layers of DNA to transmit
and receive data on leased and switched lines, and to control line turnaround operations
on half-duplex communications links. These services are used by the Data Link layer.

2.3.1.3

Call References

DNA Modem Connect uses call references to identify calls locally. Call references are
values assigned by DNA Modem Connect to outgoing call attempts placed to the network
and incoming calls which successfully connect. Each value is unique, allowing every call
(attempt) to be unambiguously identified. By identifying calls in this way, a number of
additional features are possible. In particular, call references provide:
• A means to tie together the call signaling (call establishment and release) and data
transfer phases of a call, avoiding potential problems of mis-synchronization between
the phases.

CHAPTER 2.

PHYSICAL LAYER

• A means by \vhich management information for the call signaling and data transfer

phases can be correlated by network management.
2.3.1.4

Call Sharing

DNA Modem Connect provides a feature which permits calls on a switched line to be
accessed by more than one client. This is known as call sharing. Call sharing allows, for
example, DNA MOP and the Network layer to utilize the same line. When call sharing is
used, clearing is coordinated by DNA Modem Connect such that a request by one client
to clear a call does not have immediate effect if it is still in use by others.
2.3.1.5

Network Management

In Phase V, a Line management entity :s defined which contains Physical layer management
attributes. These include speed attributes to control primary and fall-back transmission
rates, type attributes which define line and modem capabilities (null modem, switched or
leased line), a communications mode attribute denoting asynchronous or synchronous operation and an attribute to control the transmission encoding technique used, for example,
NRZI encoding.

2.4

X.21 Functional Description

X.21 is a Recommendation of the International Telegraph and Telephone Consultative
Committee (CCITT). It describes the electrical, mechanical and procedural interfaces
between Data Terminal Equipment (DTE) and Data Circuit-terminating Equipment (DCE)
for operation over synchronous public data networks.
X.21 defines two types of service: leased circuit and circuit switched. The leased circuit
service provided to higher layers is equivalent to that provided by more traditional leased
line services. In its simplest form, two DTEs attached to the network are connected pointto-point. The circuit switched service is akin to that provided by the telephone network. A
DTE attached to the network controls the establishment and release of physical connections
to other DTEs. The establishment and release procedures are defined by X.21, with the call
set-up phase involving the exchange of address and facility data with the network. Higher
layer data is carried transparently over leased and established circuit switched physical
connections.
DNA X.21 defines how the services provided by X.21 networks are integrated into DNA.
It defines the mechanisms which allow higher layers of DNA to use these services, and for
network management control and monitoring of subscriber lines. DNA X.21 supports
leased circuit and circuit switched connections.

2.4.1

DNA X.21 Features

This section describes some of the major features of DNA X.21.

2.4. X.21 FUNCTIONAL DESCRIPTION

2.4.1.1

Call Control Services

DNA X.21 defines services which permit the control and monitoring of X.21 circuit switched
connection establishment and release. This allows clients to make outgoing calls, handle
incoming calls, access call information and clear calls. Call control services do not apply
to leased circuit connections.
Clients can request the use of outgoing call facilities such as reverse charging or closed
user groups on a per-call basis. CUG and other facilities, if available from the network, may
be used to set up incoming call selection criteria which limit the incoming calls reported
to a client to those matching the criteria, for example to calls specifying a particular CUG.
Call control services allow a client to monitor the progress of an outgoing call attempt
and obtain reason information from the network if the call fails to connect. When a call
connects, the client can request full details of the call as it was actually established, for
example if it was redirected and if so, where to.

2.4.1.2

Data Transfer Services

DNA X.21 defines services which permit the higher layers of DNA to transmit and receive
data on leased circuit and circuit switched connections, and to control half-duplex operations where these are supported by the network. These services are used by the Data Link
layer.

2.4.1.3

Call References

DNA X.21 uses the call reference mechanism as described in Section 2.3.1.3.

2.4.1.4

DTE Sharing

DNA X.21 provides a feature which permits a subscriber line to be shared by more than
one client. There are two variants of this: call sharing, in which many clients are permitted
access to the same call, and line sharing, in which only one client has access to any single
call. DTE sharing allows, for exa.mple, DNA MOP and the Network layer to utilize the
same subscriber line.
Line sharing is only possible if the network provides Closed User Group and/or sub·
addressing facilities enabling different selection criteria to be set for each client. When call
sharing is used, clearing is coordinated by DNA X.21 such that a request by one client to
clear a call does not have immediate effect if it is still in use by others.

2.4.1.5

Network Management

DNA X.21 provides a comprehensive set of network management features allowing control
of subscriber lines, use of optional facilities, selection of operating procedures and the
definition of outgoing call parameter and call selection parameter profiles or templates.
DNA X.21 allows local and remote loopback testing to be performed, and on-line registration and cancellation of network facilities where this is supported by the network. This

CHAPTER 2. PHYSICAL LA YER

allows, for example, the direct call number t.o he. se.t or incoming calls to be redirected to
another DTE.
Network management counters are maintained for each subscriber line, and significant
events are logged such as DTE time-out, indicating incorrect network operation, and calls
rejected.

Chapter 3

Data Link Layer
The Data Link layer provides means to establish, maintain and release data link connections
between two directly connected systems, and for Data Link layer clients to transfer data.
Mod ules in the Data Link layer co-operate to provide this logical communications path by
use of a data link protocol. A data link protocol defines formatting rules and procedures
for data transmission and reception over a physical communications link. In the absence
of errors from the communications link, the task of the data link protocol is relatively
simple. Once errors occur, however, data corruption and loss are possible, introducing
synchronization problems between the transmitter and receiver. A data link protocol
designed to operate under these conditions has error detection and recovery procedures.
Data Link layer clients are notified of persistent or unrecoverable errors.
Data Link layer protocol modules use the services of the Physical layer described in
Chapter 2 to transmit and receive data on communications links. Physical layer services to
control operation of the link are also used. For example, when operating over a half-duplex
modem connect line, a Data Link layer module uses Physical layer services to control line
turnaround between transmit and receive modes.

3.1

Data Link Layer Functions

The following functions are performed by the Data Link layer.

Connection establishment and release. These functions are provided where the Data
Link layer module supports a connection mode type of operation. They allow synchronization of data link operation with a peer module prior to data transfer, a
process called data link connection, and termination of the connection when it is no
longer required.
An alternative mode of operation called connectionless is supported by some Data
Link layer modules. In this case, there are no establishment and release phases; the
recipient of data is determined from addressing information supplied by the client
with each service data unit.

CHAPTER 3. DATA LINK LAYER

Data transfer, User data is passed between the Data Link layer and clients in service
data units. On the transmit side: this user data is enveloped with protocol control
information before being sent. On the receive side, the control information is extracted before presentation of the data to the client. There is a one-to-one mapping
of service data unit to protocol data unit.
Framing and synchronization. The synchronization functions assemble received Physicallayer service data units (bits) into larger units (bytes and messages). The framing
function detects the beginning and end of protocol messages, once synchronization
has occurred.
Sequence control. This function uses sequence numbers in the protocol to ensure that
the sequential ordering of data transmitted on the data link connection is maintained.
It does not apply to connectionless operation.
Error detection. This function detects transmission, format and procedural errors on
the data link connection.
Error recovery. This function attempts to recover from transmission, format or procedural errors. Persistent or unrecoverable errors are reported to the Data Link layer
client. It does not apply to connectionless operation.
Flow control. This function limits the rate at which data is sent or received on the data
link connection. It does not apply to connectionless operation.
Identification and parameter exchange. This function enables the exchange of identification and operational parameters over the data link connection, typically prior
to user data transfer. This function may include parameter negotiation.
Use of Physical layer services. The Data Link layer uses the services of the Physical
layer to transmit and receive data, and where applicable, to control the operation of
the communications link.
Management. These functions allow network management to control and monitor the
operation of the Data Link layer, for example to set data link protocol operating
characteristics, enable and disable data link connections and monitor the status of
enabled connections. Data Link layer modules maintain management counters and
report significant events.

3.2

Data Link Layer Protocol Modules

Currently, the following protocol modules are defined in the DNA Data Link layer:
• Digital Data Communications Message Protocol (DDCMP). This is described in Section 3.3.
• High-Level Data Link Control (HDLC). This is described in Section 3.4.
• CSMA/ CD LAN. This is described in Chapter 4.

3.3. DDCMP FUNCTIONAL DESCRIPTION

3.3

DDCMP Functional Description

DDCMP was designed in 1974 by Digital specifically for DNA. It is a general-purpose,
byte-oriented data link protocol which is designed to serve the needs of inter-computer
data communications in a wide variety of applications and environments.
DDCMP operates over synchronous or asynchronous, switched or non-switched communications links. It can operate in point-to-point configurations or in multipoint configurations in which communication takes place between a control station and each of several
tributary stations. DDCMP messages are framed as sequences of bytes, beginning with
a single control byte indicating the message's starting point and type (data or control).
While DDCMP control messages have a fixed length, data messages have variable lengths,
indicated by a length field. On reception, this encoding allows the receiver to determine
the beginning and end of messages. The encoding also allows user data transparency;
data containing bit patterns that are those of control characters is not misinterpreted by
DDCMP. Incoming bits are assembled into bytes using start/stop bits for asynchronous
links and synchronization characters for synchronous links.
DDCMP uses two 16-bit cyclic redundancy checks (CRC-16); one to detect errors in the
protocol header, and a second to detect errors in user data. On half-duplex or multipoint
links, DDCMP executes link allocation procedures to ensure that two or more stations
do not conflict in their use of the link. These techniques are based on polling in which
one station gives permission to the other to transmit. DDCMP uses timers and sequence
numbers to detect and recover from lost messages; it also prevents the process of error
recovery from creating duplicate messages. The Network layer uses the error detection and
retry capability of DDCMP to verify that links between directly connected systems are
operational and to synchronize the operation of the routing protocols.
DDCMP makes efficient use of high speed communication links, such as satellite links,
to provide high data throughput.

3.3.1

DDCMP Features

This section describes in more detail some of the major features of DDCMP and how they
affect the operation of the protocol.

3.3.1.1

Link Management

Link Management resolves transmit contention on half-duplex point-to-point and multipoint links. In the half-duplex point-to-point case, one station transmits while the other
receives. The switching between transmit and receive modes is accomplished by a selection
flag in DDCMP messages. The transmitting station sets the flag in the last message sent,
indicating that the receiver should enter transmit mode following reception of the message.
In the multipoint case, the link appears as a party line with one station designated the
control station and all others tributaries. Each tributary is assigned an address using network management. Messages sent by the control station contain the address of a tributary.
Tributaries ignore received messages unless they are the addressee.

CHAPTER 3.

DATA LISK LAYER

Tributary stations transmit only when selected to do so by the control station. Like
the half-duplex point-to-point case, this is done using the selection flag. Messages sent
by a tributary to the control station are seen and ignored by all other tributaries. The
control station receives the message and verifies that it has come from the selected station.
Tributary stations may not exchange data directly; all traffic is between the control station
and the tributaries.

3.3.1.2

Message Exchange

The message exchange feature of DDCMP is that part of the protocol which provides
sequential, error-free data transfer. DDCMP is a positive acknowledgment retransmission
protocol. For each data message correctly received and passed to the client, a positive
acknowledgment is returned on the link to the transmitter. If incorrectly received, the
message is not passed to the client. Instead, an error recovery mechanism is invoked
causing message retransmission. DDCMP uses the CRC-16 cyclic redundancy check for
error detection on protocol headers and user data.
The basic message exchange mechanism uses data message sequence numbers, a positive
acknowledgment control message (ACK) and a timer. However, DDCMP has additional
control messages and operational techniques to achieve higher performance and faster error
recovery.

Negative acknowledgment (NAK). The duration of a timer used to detect an error
when an ACK is not returned, must be long enough to account for propagation
delay, line turnaround and processing of the data message and ACK generation by
the receiver. Typically, values of a few seconds will be used. However, DDCMP uses
a negative acknowledgment mechanism which, for most error conditions, enables
retransmission to take place before the timer expires, making more efficient use of
link resources. This mechanism is used by the receiver when it does not correctly
receive a message but does recognize that one was sent to it. Under these conditions,
a NAK control message is returned, initiating retransmission.
Reply to message number (REP). If the timer expires, this probably means that the
returned ACK was lost or corrupted. In this case, rather than retransmit perhaps
a large amount of data, a REP control message is sent containing the number of
the previous message. This forces the transmitter and receiver to synchronize their
numbering and to cause retransmission if required. If the message with that number
was received correctly, the response to REP is an ACK; otherwise it is a NAK. NAK
causes retransmission to occur.
Pipelining. Pipelining is a technique which allows further messages to be sent while
awaiting acknowledgment of ones previously transmitted. To achieve this, DDCMP
messages are numbered from 0 to 255. The numbering is cyclic {modulo 256)1 meaning that after message number 255 the next number used is O. ACKs not only
confirm that the specified message has been received correctly, but also all other unacknowledged messages since the previous ACK. If an ACK is lost or corrupted, the
acknowledgment information is automatically included in subsequent ACKs.

3.3. DDCMP FUNCTIONAL DESCRIPTION

Piggybacking. The purpose of an ACK is to convey the message number of the last successfully received data message. If data traffic is being transmitted in both directions,
this message number can be carried in a message sent in the reverse direction. This
technique is called piggybacking, and saves the framing overhead for a separate ACK
message.
ACK implied in NAK. Like an ACK, a NAK contains the message number of the last
message successfully received. Therefore, as well as indicating an error, receipt of a
NAK implies acknowledgment of previous messages.
Initialization. The process of setting message numbers to initial values is called initialization. It is accomplished by use of STRT (Start) and STACK (Start Acknowledge)
control messages. Initialization is carried out at link start-up time or after a failure,
to reset the message numbers at both ends of a data link connection to zero. The
protocol is designed so that one end of a connection cannot be initialized without the
other.
3.3.1.3

Maintenance Mode

In maintenance mode, DDCMP provides a connectionless service to the DNA Maintenance Operations Protocol (MOP). In this mode, DDCMP framing and link management
features are used but not the message exchange features described in Section 3.3.1.2. Sequence number management and message acknowledgment, if required, must be handled
by MOP itself. A data link connection cannot be used to carry normal DNA traffic while
in maintenance mode. MOP is described in Section 10.2.3.

3.3.1.4

Network Management

The operation of DDCMP is unchanged from Phase IV. Management of the DDCMP
module, although largely unchanged in function, has been upgraded in accordance with
the Phase V network management model. In Phase IV, the Line database contained both
Data Link layer protocol and Physical layer management parameters. In Phase V, management of Data Link layer protocols is separated from Physical layer management. For
each communications link on which DDCMP is to operate, a Link management entity is
created. This contains the data link management attributes common across all DDCMP
stations on the line, such as duplex mode and protocol type. Additionally, each Link entity
has one or more Station management sub-entities, where each Station contains data link
management attributes specific to operation with a single DDCMP station on the line, such
as station address and protocol state. In point-to-point and multipoint tributary configurations, each Link entity has a single Station sub-entity. In multipoint control configurations
each Link entity has up to 255 Station sub-entities; one for each tributary controlled by
the protocol.
The Link and Station management entities contain the data link protocol related attributes which in Phase IV were held in the Line and Circuit databases respectively.
The I-bit DDCMP counters of Phase IV have been replaced in Phase V by larger
counters. Amore comprehensive set of management events has been defined.

CHAPTER 3. DATA LINK LAYER

:1.4

HDLC Functional Description

HDLC is a bit-oriented data link protocol which operates over synchronous, switched or
non-switched communication links. Its various aspects are described in a series of standards
including:
• ISO 3309 -

HDLC frame structure

• ISO 4335 -

HDLC elements of procedure

• ISO 7809 -

HDLC classes of procedure

• ISO 8471 -

HDLC data link address resolution

• ISO 8885 -

HDLC general purpose XID information field content and format.

In addition, a subset of HDLC known as LAPB, which is used in X.25 level 2, is defined
m:

• ISO 7776 -

X.25 DTE data link layer

• CCITT Recommendation X.25 (1984)
Taken together, these standards define a whole range of protocols, procedures and features,
many of which are not capable of inter-operation. The DNA specification of HDLC uses a
subset of these, selected to provide the characteristics and features required by DNA and
to permit inter-operation with other common subsets.

3.4.1

HDLC Features

The following sections describe in more detail, some of the major optional features of HDLC
and the way they are used in DNA HDLC.
Operational modes. The HDLC standards specify various operational modes. In DNA
HDLC two modes are supported. Normal Mode is used only over a half-duplex
physical line, whereas the more efficient Balanced Mode is used over a full-duplex
line. A DNA HDLC station in Normal Mode can operate as a primary on a point to
point physical link, or as a secondary on either a point to point or multipoint physical
link.
Extended sequence numbering. In the base HDLC standard, sequence numbers are
modulo seven, but when the optional extended sequence numbering is in use, this
is increased to 127. This is required for efficient operation over physical links which
have a large propagation delay in relation to their bandwidth (notably satellite links).
DNA HDLC will normally be configured to use extended sequence numbering.
Error detection. HDLC uses a single cyclic redundancy check to detect errors in both the
header and data fields. This may be either a 16-bit or 32-bit CRC. DNA HDLC uses
a 32-bit CRC by default in order to provide maximum protection against undetected

3.4. HDLC FUNCTIONAL DESCRIPTION

errors, but it will use a 16-bit CRC when communicating with a station which does
not support the use of a 32-bit CRe. Additionally, when HDLC is being used between
two DNA stations, a user data length field is included in the user data.

Negotiation of options. Many of the operational parameters of HDLC, such as station
address, window size or the use of various options, must be set correctly before two
stations can communicate. A special frame, the XID frame, may be exchanged during
link initialization to allow these to be negotiated to mutually acceptable settings.
Support for the XID negotiation procedures is included in DNA, but it is defined to
be optional by the ISO standards. When communicating with a remote non-DNA
station which does not support XID negotiation, the stations must be manually
configured via network management.
Maintenance operations. Unlike DDCMP, there is no special Maintenance Mode in
HDLC. Such messages are instead carried in UI (Unnumbered Information) frames.
These may be exchanged even when the sequenced data mode has not yet been
initialized. Since UI frames are not subject to error correction, sequencing or flow
control, such functions are handled in DNA by the higher layer management protocol
(MOP).

CHAPTER 3. DATA LINK LAYER

Chapter 4

Local Area Networks
Local Area Networks (LANs) are a type of communication facility that provides high
speed communication in a moderate size geographical area. Phase V of DNA supports
the CSMA/CD (Carrier Sense Multiple Access with Collision Detect) class of LAN, as
defined by the ISO 8802-3 and Ethernet standards. In the ISO Reference Model, Local
Area Networks reside in the Physical and Data Link layers. For example, the ISO 8802-3
standard defines both Physical layer aspects such as the type of cables to be used, and
Data Link layer aspects such as frame formats and services provided to higher layers.

4.1

Services Provided

The DNA CSMA/CD Data Link module provides the following services:

Transmission and reception of datagrams. The LAN Data Link module detects data
corruption, but does not guarantee delivery of all frames. This type of Data Link
service is a datagram service.
Support for multiple stations. A LAN may have thousands of stations; each has a
unique address. Datagrams specify both source and destination addresses.
Multicast. Datagrams may be sent to a multicast (group) address to send to a group of
related stations with a single frame.
Multiple protocols. The LAN Data Link module allows for multiple protocols to be used
on the same LAN, and on the same station.
Shared access. Because the LAN is shared among all the attached stations, each station
controls its access to the LAN to ensure fair and efficient use of the shared resource.
Management. The LAN Data Link module provides a management interface for monitoring of usage, diagnosis and control.
29

CHAPTER 4.

4.2

LOCAL AREA NETTA'ORKS

Characteristics of the CSMA/CD LAN

Some of the characteristics of the CSMA/CD LAN are:
• Physical layer using coaxial cable segments of up to 500 meters length, or Thin Wire ™
cable segments of up to 185 meters length.
• Segments may be interconnected by repeaters into a LAN. The resulting LAN must
have a tree topology (without loops), and there may be at most two repeaters between
any two stations. There may be up to 1024 stations on a LAN, with a maximum
distance between any two stations of 2.8 kilometers.
• Channel bit rate of 10 million bits per second.
• Frame size in the range 64 to 1518 bytes.
• Sharing of access by carrier sense and collision detection. Carrier sense means each
station waits for the channel to become quiet before transmitting. If two stations
start transmitting simultaneously, this produces a collision, which is detected and
causes the transmission to be retried as needed. The result is a Data Link which
offers very low delay at low to moderate loads, and degrades gracefully under high
loads or overloads, without instability. With this simple mechanism, the CSMA/CD
LAN provides fair service to all stations regardless of load.
• LANs may be interconnected by bridges, thereby creating an Extended LAN. Refer
to Section 4.5 for more detail.

4.3

Relationship to International Standards

The ISO 8802 series of standards define a number of Local Area Networks. (These standards
were originally developed by IEEE; they have since become ISO standards.) Phase V DNA
supports the CSMA/CD class of LAN, as specified by ISO 8802-3 (IEEE802.3). It also
supports the Ethernet standard.
LAN Bridges are the subject of IEEE 802.1, part D (MAC Bridges) and ISO 8802-1.
Significant portions of these standards are based on the architecture of DNA Bridges. Since
these standards are still in a draft stage, conformance to the eventual standards is not yet
assured but migration will be greatly eased.

4.3.1

Addressing

Local Area Networks support connection of multiple stations, so each frame has to specify
which station(s) should receive it. The ISO 8802 standards specify 48-bit addresses for this
purpose. Each frame contains both a destination address and a source address.
The first two bits in the address (Figure 4.1) have a special function. The first bit is the
Individual/Group address bit. If this bit is set, the address is a group (multicast) address.

4.3. RELATIONSHIP TO INTERNATIONAL STANDARDS

bits

IGILI: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : I
Figure 4.1: LAN address format
If this bit is clear, the address is an individual station address. The source address must
be an individual address.
The second bit is the Globally/Locally administered address bit. If this bit is clear,
the address is from an address block assigned by the IEEE. If this bit is set, the address is
from a locally managed address block. For example, addresses beginning with the 24-bit
block identifier 08-00-2B are assigned to Digital, and Digital assigns individual unique
addresses from that block to its products. When locally managed addresses are used, it is
the responsibility of the local administration to ensure that addresses are unique.
Multicast addresses are used to send a single copy of a frame to a number of stations.
The multicast address designates a particular class of station (and LAN Data Link user)
rather than a specific station. For example, the Network layer on DNA Level 2 routers
receives frames addressed to the multicast address 09-00-2B-02-00-00 (see Section 5.2.2).

Multiplexing on ISO 8802-3 LAN s

4.3.2

The LAN Data Link module allows multiple Data Link users on the same system (multiple
higher level protocols) to access the LAN concurrently. This is done using the procedures
and frame formats defined in ISO 8802-2 (IEEE 802.2), Logical Link Control (LLC). DNA
uses LLC Class 1, which provides a datagram service. The standard also defines Class
2, which is not used by DNA protocols. The general format of an LLC frame on the
ISO 8802-3 LAN is shown in Figure 4.2.

I Pre I aest

1-2

Source

Length

aSAP

SSAP

etl

bytes

I user::aata IPad IeRe

Figure 4.2: ISO 8802-3 frame format
Description of the fields:
Pre. Preamble and starting delimiter. This is the bit pattern 1010 ... 1011, which the
CSMA/CD Physical layer uses to locate the start of the frame.
Dest. 48-bit destination address.
Source. 48-bit source address.

CHAPTER 4.

LOCAL AREA NETWORKS

Length. The number of bytes in the DSAP, SSAP, Ctl and User Data fields.
DSAP. Destination Service Access Point address.
SSAP. Source Service Access Point address.
Ctl. LLC Control field. For LLC Class 1 this is one of:
UI. Unnumbered Information (datagram) LLC frame.
XID. Exchange Identification LLC frame.
Test. Test (loopback) LLC frame.
User Data. Data supplied by the LAN Data Link user.
Pad. Pad bytes used, if necessary, to fill out the frame to a minimum total size of 64 bytes.
CRC. The ISO 8802-3 32-bit Cyclical Redundancy Check error detecting code.
In the same way that destination and source address identify the receiving and sending
station, the destination and source SAP (DSAP and SSAP) addresses identify the receiving
and sending LAN Data Link user. The values for the SAP address fields are assigned by
the IEEE; each value identifies a particular protocol defined by a national or international
standard. For example, the SAP address assigned to the ISO Connectionless-mode Network
protocol (ISO 8473) is FE (01111111).
Since SAP addresses are assigned only to standardized protocols, another frame format
is used for protocols that are not national or international standards. This frame format
is known as Subnetwork Access Protocol ( SNAP) and is illustrated in Figure 4.3. It is a
variant of the UI frame of LLC Class 1, with SSAP and DSAP both set to the SAP address
for SNAP, which is AA.
The five bytes following the Ctl field are the Protocol ID, which is used to designate
a particular LAN Data Link user. The first three bytes are taken from the 24-bit block
identifier assigned by IEEE (see Section 4.3.1); the last two bytes are assigned by the
owner of the block. For example, 08-00-2B-80-3C is the Protocol ID for the DNA Naming
Service, advertising availability of Name Servers on a LAN. As with station addresses, the
first two bits of the Protocol ID have special significance. The first bit is reserved and
must be zero. The second bit is zero for globally assigned Protocol IDs, and one for locally
assigned Protocol IDs.

Pre

Dest

Source

Length

Prot ID

bytes
Use rData

Figure 4.3: ISO 8802-2 SNAP frame format

I Pad ICRC I

4.4. ETHERNET

4.4

Ethernet

Ethernet is an earlier standard for a CSMA/CD LAN very similar to ISO 8802-3. It
uses the same Physical layer as ISO 8802-3, and the same format 48-bit addresses and
CRC. Multiple protocols are supported differently than in LLC. Figure 4.4 illustrates the
Ethernet frame format.
8

I Pre I Dest

Source

bytes

I Prot Type I user::Data IeRe I

Figure 4.4: Ethernet frame format
The Ethernet frame format uses the two-byte Protocol Type field to identify the LAN
Data Link user. The Protocol Type values are assigned by the Ethernet address administration. For example, 90-00 is the protocol type for the Ethernet loopback protocol.
Unlike the ISO 8802-3 frame format, the basic Ethernet frame format does not include
a length field. The CSMA/CD Physical layer requires that all frames must be at least 64
bytes long (not including the Pre field). In the basic Ethernet frame format, the LAN Data
Link user is required to ensure this by always supplying at least 46 bytes of user data for
each frame (attempts to supply less data are rejected by the LAN Data Link module).
To eliminate the need for each Data Link user to deal with the minimum CSMA/CD
frame size, the DNA CSMA/CD Data Link module also supports a variant of the basic
Ethernet frame format, which includes a length field and padding. This format is shown
in Figure 4.5. There is no indication in the frame header to distinguish the padded format
from the basic format; it is up to the communicating LAN Data Link users to agree (for
any particular Protocol Type) whether the padded form or the basic form will be used.

Pre

Dest

Source

bytes

I Prot Type I Length I user::Data I Pad IeRe I

Figure 4.5: Ethernet frame format with padding

4.5

Bridges

Local Area Networks may be interconnected by LAN Bridges to create Extended Local
A rea Networks. A Bridge performs a Data Link layer relay function which is transparent

CHAPTER 4.

LOCAL AREA NET~rORKS

Lo die Dala Link users~ and independent of the protocols used above the Data Link layer.
Bridges handle both ISO 8802-3 and Ethernet frames.

4.5.1

Bridge Services

The purpose of Bridges is to extend the LAN transparently, preserving the services available in the basic LAI\ (see Section 4.1) while increasing the number of stations allowed,
the maximum distance between stations, and the total available bandwidth. The Bridge
functions are layered on top of the LAN Data Link layer module. The constraints on LAN
size (total number of stations, number of stations) are constraints of the layers below the
Bridge functions; therefore they apply only to the individual LANs in a Extended LAK,
not to the Extended LAN as a whole. Note that the Bridge uses the LAN Data Link layer
module in such a way that it processes all frames, regardless of whether they are ISO 8802-3
or Ethernet format.
The bandwidth of a Extended LAN is greater than that of a LAN because Bridges only
forward frames when needed. If the source and destination for a particular frame are on
the same side of a Bridge, it will not forward the frame; other LANs in the Extended LAN
can thus carry other frames at the same time.

4.5.2

Extended LAN Topology

Local Area Networks may be interconnected by Bridges with few topological constraints.
Topologies with loops are permitted; in general the Extended LAN may have an arbitrary
mesh topology. The constraints that do apply concern the total number of stations in
the Extended LAN, and the number of Bridges in the path between any two stations
(the network diameter). The total number of stations is limited by the capacity of the
Forwarding database in the particular bridge implementation used; the network diameter
is limited by the need to limit the total delay across the Extended LAN.
The advantage of this flexibility is that it allows topologies with redundancy. For
example, a pair of LANs could be connected by two Bridges in parallel, or a number of
LANs could be connected in a circular topology. In either case, the Extended LAN would
remain intact even if one of the Bridges fails. More highly redundant topologies are also
possible.

4.5.3

Bridge Algorithms

4.5.3.1

Spanning Tree Computation

As described above, the topology of an Extended LAN is relatively unconstrained and may
include loops. Bridges compute a spanning tree for the Extended LAN to ensure frames
will not loop continuously. Only Bridge ports that are in the spanning tree are used for
frame forwarding.
The Bridges operate a protocol between them to exchange the information required to
compute the spanning tree. Whenever the topology changes, the spanning tree computation

4.6. ISO 8802-3 AND ETHERNET COEXISTENCE

produces a new spanning tree. For example, if two bridges are in parallel, one will be in
standby mode. Should the active bridge fail, the one in standby mode will take over.

4.5.3.2

Station Address Learning

Because Bridges forward packets using only ports that are in the spanning tree, there is
only one path from a Bridge to any given destination station. This allows the Bridge to
learn about stations by observing source addresses in received traffic, and forwarding based
on the information learned.
Bridges maintain a local forwarding database, which contains station addresses and associated Bridge ports. Whenever an error free frame is received on a port, the Bridge enters
the source address of the frame into the station address database, with the identification
of the port on which the frame was received.

4.5.3.3

Forwarding algorithm

Whenever an error free frame is received on a Bridge port, the forwarding algorithm is
used to determine how to process the frame. First, the Bridge checks whether the port on
which the frame was received is in the spanning tree. If not, then this port is currently
only a backup, and the frame is discarded. Next, the destination address is looked up in
the forwarding database. If the port ID in the database is that of the port on which the
frame was received, the Bridge simply discards the frame (since this indicates that the
source and destination are "on the same side" of this Bridge). Otherwise, it forwards the
frame on the indicated port.
H the destination address is not found in the forwarding database, this means the Bridge
has no information about the location of the destination. In that case the frame is flooded
along the spanning tree. The frame is forwarded on all ports in the spanning tree except
the one on which the frame was received. The result of the flooding is that the frame is
sent once on each LAN of the Extended LAN.
In short, Bridges learn from source addresses and forward based on destination addresses.

4.5.4

Bridge Management

Bridges are self-initializing and autoconfiguring. That is, they will operate correctly without requiring any management intervention.
Management facilities are provided to allow observation and diagnosis. They can also
be used to make permanent entries in the forwarding database, for example to prevent
certain traffic from being forwarded at all.

4.6

ISO 8802-3 and Ethernet Coexistence

DNA allows both ISO 8802-3 and Ethernet format frames to be used on the same LAN,
and even on the same system. Each Data Link user specifies to the LAN Data Link module

CHAPTER 4.

LOCAL AREA NETWORKS

which format is to be used (ISO 8802-3 LLC, ISO 8802-3 S::\AP, basic Ethernet, Ethernet
with padding).
The protocol frames can be distinguished by examining the two byte field following the
Source Address. In ISO 8802-3 format, this is the length of the LLC data; in Ethernet
format, the Protocol Type. The minimum size LLC frame has one byte DSAP, SSAP and
Ctl fields and a null User Data field, and the maximum CSMA/CD frame size (not counting
the preamble) is 1518 bytes; therefore the Length field in the ISO 8802-3 format will be in
the range 3-1500.
When the CSMA/CD Data Link module receives a frame, it examines the two byte
field following Source. If this field, interpreted as a two byte integer with the high order
byte coming first, is in the range 3-1500, the frame is an ISO 8802-3 frame. Otherwise it
is an Ethernet frame. As a result, the Ethernet Protocol Type codes in the range 00-03
to 05- DC are unavailable, since those values in the Prot Type field would be interpreted
as ISO 8802-3 lengths. These values are therefore not assigned to any Ethernet Protocol
Type.
DNA Phase V provides Ethernet support for compatibility with Phase IV and with
other existing protocols that use Ethernet format frames. For new applications, the
ISO 8802-3 format is preferred because it is an international standard, and also because
it provides some LAN technology independence. An application using ISO 8802-3 could
easily be migrated to other LANs in the ISO 8802 family, since the same set of services is
available on all of them.
If an existing application that uses the LAN Data Link module needs to migrate to
ISO 8802-3 format frames, some temporary duplication of function will be necessary during the migration to permit the application to interoperate with existing (Ethernet-only)
versions. The new version would use ISO 8802-3 frame format as the default format, but
would also listen for frames in Ethernet format. If a system using the older version attempts to communicate with the new version, the new version will notice that the request
was made in Ethernet format and respond in the same format. For example, the DNA
Phase V Network layer uses ISO Connectionless-mode Network protocol (ISO 8473) using
ISO 8802-3 frame format, but also listens to Phase IV frames in Ethernet format. If it
receives any, it records the fact that the sending system is a Phase IV system and sends
appropriate responses in Phase IV format using Ethernet frames.

Chapter 5

Network Layer
The Network layer of DNA provides the ISO Connectionless-mode Network Service (CLNS)
defined in ISO 8348 Addendum 1, using the protocol defined in ISO 8473. It also provides
the ISO Connection-mode Network Service (CONS) defined in ISO 8348. The provision of
CLNS is described below, while the provision of CONS is described in Chapter 9.
The Network layer accepts messages (NSDUs) from the local Transport layer modules
via a local Network Service Access Point (NSAP) and delivers them to destination Transport layer modules via remote NSAPs. The particular destination NSAP is identified by
means of its NSAP address. The information in the NSDU is transmitted through the
network in one or more data Network layer Protocol Data Units (NPDUs). The Network
layer is responsible for forwarding the data NPDUs, possibly through multiple intermediate
systems, to the destination NSAP. In order to achieve this, it must be able to determine
the best path (or paths) to each destination, based on the current topology of the network.
The network topology comprises a (possibly very large) number of systems interconnected by various subnetworks (such as DDCMP or HDLC point to point links, ISO 8802-3
LANs or X.25 packet switched networks). This topology will be constantly changing, both
as a result of new systems and their connecting subnetworks being added, and the failure
and subsequent repair of existing ones. In order to determine the best paths in the face
of these perturbations, the DNA Network layer employs a distributed Routing Algorithm
based upon the shared knowledge of all systems within a DNA Routing Domain (Routing Domains are described in Section 5.2.1). Each system is responsible for ascertaining
the reachability and identity of its immediate neighbors (that is, those systems currently
reachable by traversal of a single subnetwork). The information about the state of each
system's links, together with a locally assigned cost for each link, is used to construct a
Link State PDU (LSP) for that system. LSPs are propagated periodically (and whenever
there is a change in any of the links to that system) throughout the routing domain.
As a result of receiving LSPs from all the other systems, each system is in possession
of information about every other system's links. This information is used by each system
to construct a representation of the current network topology and hence compute the best
(that is, least cost) path to each particular destination.
37

CHAPTF:R S

5.1

NPTVVORK LA YER

Network Service Characteristics

The service provided by the Connectionless-mode Network Service has the following characteristics.

• It is a datagram service. Messages are delivered to their destination on a best-effort
basis.
• Messages of up to 65535 bytes are accepted for transmission.
• There exists the possibility that the Network Service will:
Duplicate a message.
Deliver a sequence of messages to a given destination in a different order to that
in which they were presented at the source.
Fail to deliver a message.
• There is an extremely low probability that the Network Service will deliver a message
to the wrong destination or modify the data in a message.
• There is a variable delay time between the presentation of a message at the source
and its subsequent delivery at the destination.

5.2

Network Layer Concepts

The following basic concepts are essential to an understanding of the operation of the
Network layer.

5.2.1

Hierarchical Routing

A collection of End Systems, Intermediate Systems and Subnetworks operated by a single
organization or administrative authority is known as an Administrative Domain. An Administrative Domain may be subdivided into a number of Routing Domains, some or all
of which may be DNA Routing Domains.
A Routing Domain is a set of End Systems, Intermediate Systems and Subnetworks
which operate according to the same routing procedures and which is wholly contained
within a single Administrative Domain.
In order to support very large domains, possibly containing several hundred thousand
systems, Routing Domains in Phase V DNA are themselves hierarchical. A large DNA
Routing Domain may be partitioned by the network manager into sub-domains known as
areas. Each system resides in exactly one area. Routing within an area is referred to as
Level 1 routing. Routing between areas is referred to as Level 2 routing. (Routing between
Routing Domains is performed by means of static tables as described in Section 5.2.5.)
Level 2 routers keep track of the paths between areas. Levell routers keep track of the
routing within their own area. For an NPDU destined to another area, a Level 1 router
sends the NPDU to the nearest Level 2 router in its own area, regardless of the actual

5.2. NETWORK LAYER CONCEPTS

destination area. The NPDU travels via Level 2 routing to the destination area, where it
again travels via Levell routing to the destination system.
The division of a large DNA Routing Domain into separate areas improves network
performance by drastically reducing the amount of routing overhead compared to a single
area of the same size. It also allows the inter-area traffic to be confined to a particular set
of systems and subnetworks.

5.2.2

System Types

The following types of system exist in DNA.

End Systems. These systems originate NPDUs for transmission to other systems and
receive NPDUs from other systems, but do not relay NPDUs. In general End Systems only have one attached subnetwork. Multilink End Systems however, may have
multiple attachments to one or more subnetworks (for example, to provide greater
resilience to failures).

Intermediate Systems. These systems may originate and receive NPDUs from other
systems, but also relay NPDUs from other source systems to other destination systems. In DNA, such systems are termed routers. There are two types of router.

Levell Routers. These systems route directly towards systems within their own
area, and route towards a Level 2 router when the destination system is in a
different area.

Level 2 Routers. These systems act as Levell routers within their own area, but
in addition route to other areas. They route to other routing domains by means
of static routing as described in Section 5.2.5.

5.2.3

Addressing

The structure of Network layer addresses within the global OSI environment is defined
in ISO 8348 Addendum 2. This introduces the concept of addressing domains. At the
highest level is the global network addressing domain which consists of all NSAP addresses
in the OSI environment. This is subdivided into a number of sub-domains each of which
is associated with an addressing authority. The authority may itself allocate addresses
within its domain, or it may further subdivide its domain and assign an authority to each
of its sub-domains. This process may be continued to an arbitrary extent, limited only
by the maximum NSAP address length. The uniqueness of addresses within a particular
hierarchical domain is ensured by the authority responsible for that domain.
The address format defined by ISO is illustrated in Figure 5.1. It divides an address
into two major parts, the Initial Domain Part (IDP) and the Domain Specific Part (DSP).
The IDP is further divided into an Authority and Format Identifier (AFI) and the Initial
Domain Identifier (IDI).
The AFI specifies the format of the IDI and the authority responsible for allocating its
values. ISO has specified a number of such authorities, among which are:

CHAPTER 5. NETWORK LA YER

+'4----NSAP address - - - -....

I AFI I

IDI

DSP

IDP-----

Figure 5.1: NSAP address structure
• X.121, the CCITT numbering plan for X.21 and X.25 networks.
• E.163, the CCITT numbering plan for the telephone network.
• E.164, the CCITT numbering plan to be used for ISDN networks.
• ISO Geographic. This is an ISO defined scheme which allows a country to assign
addressing domains independently of public networks.
• ISO Non-geographic. This is an ISO scheme which allows international organizations
to assign domains which do not imply the location of systems within the domain.
NSAP addresses in Phase V DNA may be constructed with any valid AFI and IDI.
To take a specific example, in the case of an AFI which specifies the X.121 numbering
plan, the IDI is an X.121 address. This would typically, but not necessarily, be the address
of the DTE by which a group of private subnetworks are attached to the public network.
Although it is useful to think of a correspondence between an addressing domain and
the physical topology of the network, it must be remembered that this is by no means
mandatory. In this example the X.121 address need not correspond to any physical DTE
at all. However it simplifies the routing between routing domains if such a correspondence
can be achieved, since the DTE address to be called in order to reach a particular NSAP
can then be derived from the NSAP address.
The format of the DSP is not subject to standardization and it is possible for different
structures to be used in different routing domains, since it only need be interpreted within
the defining domain. Within a DNA routing domain the DSP portion is structured as
shown in Figure 5.2 in order to facilitate the operation of the routing functions.
No. of bytes

~l-LO-C---A-R-E-A~-----------ID------------~SE-L~

Figure 5.2: Structure of the DSP in DNA
LOC-AREA. A two-byte field which is unique within an IDP. The concatenation of lOP
and LOC-AREA define the bounds of an area and is known as the A rea A ddress. If

5.2. NETWORK LAYER CONCEPTS

the Area Address of an NSAP address exactly matches the Area Address of a system,
it is in that system's area, and is routed by Levell routing. Otherwise it is routed
via Level 2 routing.

ID. A six-byte field which uniquely identifies the particular system within an area.
SEL. A one-byte field which acts as an NSAP selector for the module within a particular
system which is to receive the data NPDU. This may be a Transport layer recipient
or the Network Routing Module itself. The selector is always the last byte of the
address.
In some circumstances, it may be desirable for an area (and hence the systems within
it) to have more than one area address. For example, if the area has multiple points of
attachment to public networks, it may be useful to have NSAP addresses which correspond
to each one. However the routers making up an area must have at least one area address
in common with each of their neighbors.
In routers, the area address(es) are set by Network Management. However, End Systems can autoconfigure their area addresses to be those of their neighboring router. Each
DNA system has a unique 6 byte identification (known as the NodeID). This is used as the
ID field of the NSAP address. The SEL field is specified by the Transport layer modules.
End System NSAP addresses can therefore be completely autoconfigured without Network
Management intervention.

5.2.4

Subnetwork Types

There are two generic types of subnetworks supported by the Network layer.

Broadcast Subnetwork. These are multiaccess subnetworks that support the capability of addressing a group of attached systems with a single NPDU, for instance
ISO 8802-3 LANs (see Chapter 4).
General Topology Subnetworks. These are non-broadcast subnetworks, which may
themselves be subdivided into two types:
Permanent Point-to-point Links. These are links that stay connected at all times
(unless failed, or turned off by Network Management), such as leased lines or
private links operating DDCMP or HDLC Data Link layer protocols.
Dynamically Established Data Links (DEDs). These are links over connection
oriented facilities such as X.25 networks. They can be used in one of three ways:
Static Point-to-point (Static). The call is established upon Network Management action and cleared only on Network Management action (or failure).
Dynamic Connection Management (DCM). The call is established upon
receipt of traffic, and brought down on timer expiration when idle.

CHAPTER 5. NETWORK LA YER

Dynamically Assigned (DA). Uses dynamic connection management, but
in addition, the address to which the call is to be established is determined
dynamically from the destination NSAP address. Routing information is
not exchanged over these links, but data traffic is forwarded according to
statically entered routing information as described below.

5.2.5

Multiple Routing Domains - Static Routing

Systems within a DNA Routing Domain are able to communicate with OSI systems in other
Routing Domains. These may be in other Administrative Domains, or separate Routing
Domains within the same Administrative Domain. The latter may be desirable, in certain
cases, for reasons of:

Cost. Only data traffic need be exchanged between the Routing Domains so no additional
costs are incurred for routing control traffic.
Scale. With a two level hierarchy, there is a limit (albeit very large) to the size of Routing Domain the routing algorithm can handle. As long as each individual Routing
Domain is reasonably sized, the total size of the Administrative Domain is for all
practical purposes unbounded.
Robustness. Routing Domains are isolated from each other, and therefore protected from
such conditions as faulty routers or excessively unstable topologies that might occur
in a different Routing Domain.
Interoperability with other routing algorithms. The routing algorithm used in one
Routing Domain does not need to be the same as the algorithm used within a different
Routing Domain. Thus a DNA domain may communicate with systems in a nonDNA OSI domain.
In order to facilitate the construction of such multi-domain topologies, and permit data
to be routed between domains, provision is made for the entering of static inter-domain
routing information. This information is provided by a set of Address Prefixes entered
by Network Management at the boundary Level 2 Router. The prefix indicates that any
NSAP whose address matches the prefix may be reachable over the subnetwork with which
the prefix is associated. The prefix also has associated with it the required subnetwork
addressing information. In some cases, where the address of the subnetwork point of
attachment has been used as the IDI in the NSAP addresses matched by the prefix, it is
possible to derive the subnetwork address from the particular NSAP address. For example,
an X.121 DTE address may be obtained from the IDI of the NSAP address. This permits
a single static entry to be used to provide routing information for a number of different
points of attachment.
The Address Prefixes are handled by the Level 2 routing algorithm in the same way as
information about a Levell Area within the domain. NPDUs with a destination address
matching a prefix present on any Level 2 router within the domain can therefore be relayed
(using Level 2 routing) to that Level 2 router and delivered out of the domain. It is then

5.3. NETWORK LAYER FUNCTIONS

assumed that the routing functions of the other domain will then be able to deliver the
NPDU to its destination.

5.3

Network Layer Functions

The Network layer Routing functions may be divided into two groups:
• Subnetwork Independent Functions
• Subnetwork Dependent Functions

5.3.1

Subnetwork Independent Functions

The Subnetwork Independent Functions are independent of the specific Data Link layer
module operating below them. The following components may be identified.

Routing. The routing component determines NPDU paths. A path is the sequence of
connected systems and links between a source End System and a destination End
System. The routing component has the following specific functions:
• It extracts and interprets the routing PCI from received NPDUs.

• It performs NPDU forwarding based on the destination NSAP address.
• It manages the characteristics of the path. If a system or link fails on a path, it
finds an alternative route.
• It interfaces with the Subnetwork Dependent Functions to receive reports concerning changes in the availability of neighboring systems.

• It returns error reports to the source where necessary.

Segmentation and reassembly. Where NPDUs are too large to be transmitted in a
single Data Link layer SDU they are segmented. NPDUs are reassembled at the
destination.
PDU lifetime control. This bounds the amount of time that a data NPDU can exist in
the network, thus preventing NPDUs from circulating indefinitely. Each data NPDU
has a lifetime field which is decremented every time the NPDU is forwarded. If the
lifetime reaches zero, the NPDU is discarded and, if requested, an error report is
returned to the source.
Congestion avoidance. This manages the buffer resources used at each intermediate
system. When the average value of the queue of NPDUs for a link reaches a threshold, a congestion experienced bit is set in any NPDUs relayed over that link. This
information is used by the Transport layer congestion avoidance functions described
in Section 6.3.2.3. If the queue continues to grow, NPDUs are discarded to prevent
further congestion. In addition the congestion avoidance function regulates the ratio
of traffic being relayed by a system to traffic originated by that system.

CHAPTER 5. NETWORK LAYER

5.3.2

Subnetwork Dependent Functions

The Subnetwork Dependent Functions mask the characteristics of the Data Link layer from
the Subnetwork Independent Functions. They perform the following functions:
• Identifying whether the neighbor is a Phase IV or a Phase V system.
• Operation of the ISO End System to Intermediate System Routing Protocol (ISO 9542)
on the particular subnetwork, in order to:
Ascertain the subnetwork addresses of operational neighboring systems, and the
NSAP addresses present.
- Ascertain the subnetwork address of operational routers.
Determine the Area Address to be used for the autoconfiguration of End System
NSAP addresses.
• Operation of the requisite Subnetwork Dependent Convergence Function as defined
in ISO 8473 Addendum 1, in order to perform:
Data Link initialization.
Hop by hop segmentation over subnetworks with small maximum SDU sizes. l
Call setup and tear-down on Dynamically Established Data Links.

5.4

Routing Operation

The Routing component contains three major processes.
• Decision
• Update
• Forwarding

5.4.1

The Decision Process

This process uses the database of link state information, derived from the LSPs received
from other systems, to determine the least cost path(s) to each router in the routing domain. This information is entered into the forwarding database, from which the Forwarding
Process can determine the proper next hop for each data NPDU.
lFor example, the use of the M-bit procedures in X.25 as described in Section 9.3.1.3.

5.4. ROUTING OPERATION

5.4.2

The Update Process

This process is responsible for generating and reliably propagating LSPs around the domain. LSPs are generated periodically and also when notified by the Subnetwork Dependent Functions that a change has occurred in the reachability of the system's neighbors.
LSPs are propagated by flooding. Each new LSP is transmitted to all the router's neighbor
routers except the neighbor router from which it received the LSP.
The Update process is responsible for determining, given a received LSP, whether that
LSP represents new, old or duplicate information with respect to what is already stored in
the database and acting accordingly.
• If the LSP is new, it is written into the database, acknowledged as described below)
and forwarded on all links other than the one on which it was received.
• If the LSP is old, the stored, newer, LSP is transmitted on the link over which the
older one was received.
• If the LSP is a duplicate, it is acknowledged, but is not forwarded.

Each LSP generated by a particular router is assigned a sequence number. An LSP
with a preceding sequence number is defined to be older. Additionally, each LSP has an
age field. When the age expires the LSP is deleted from the database.
The update process is responsible for ensuring that the latest LSPs reach every reachable router in the sub-domain. This is done by sending acknowledgments. On nonbroadcast links these are sent whenever an LSP which is newer or a duplicate is received.
Older LSPs are acknowledged by sending the newer LSP. On broadcast links a special
Sequence Numbers PDU (SNP) is periodically multicast to all routers.

5.4.3

The Forwarding Process

This process inspects the destination NSAP address of a received data NPDU and determines, by reference to the forwarding database generated by the decision process, the
correct link over which to forward the NPDU. It then queues it for transmission over that
link. Where the forwarding database indicates that there are multiple equal cost paths to
the destination, the forwarding process performs load splitting. Successive data NPDUs
are transmitted on each of the indicated links in turn.
When the forwarding process is unable to deliver a data NPDU it is discarded. If the
original NPDU so requested, an error NPDU is returned to the original source, specifying
the system where the error occurred and the nature of the error.
When the forwarding process forwards a data NPDU onto the same subnetwork from
which it was received, it additionally sends a Redirect NPDU (as defined in ISO 9542) to
the originator of the data NPDU in order to inform it of the better route. Subsequent data
NPDUs can then be sent directly to the system identified in the Redirect NPDU without
further involvement of the router.

CHAPTER 5. NET\VORK LAYER

Chapter 6

Transport Layer
The Transport layer in DNA provides reliable transfer of data between Transport service
users. It operates over the Network layer and can use either style of Network service:
connectionless-mode (Chapter 5) or connection-mode (Chapter 9). The Transport layer
employs end-to-end protocols which enhance the inherent characteristics of the underlying
Network service.

6.1

Transport Layer Functions

The DNA Transport layer performs the functions described below. DNA can operate
two different Transport layer protocols (Section 6.2). The exact functions performed will
depend upon the protocol being used.

Construction and destruction of Transport Connections. A Transport Connection
is a virtual connection at the Transport layer between two Transport service users,
either on separate systems or the same system in a network. A pair of Transport
service users may have more than one Transport Connection between them. Each
Transport Connection in a network exists independently of any other. A Transport
Connection must be established before two Transport service users can exchange
data. The process of connection establishment also permits the communicating peers
to negotiate mutually acceptable characteristics for the connection.
Data transfer using flow control. Transport provides a full duplex data path between
a pair of communicating Transport service users. It transfers data from transmit
buffers in one system to receive buffers in another. Flow control mechanisms are
employed to balance the relative speeds of the transmitter and receiver. Transport
recovers from message loss or duplication occuring in lower layers and ensures that
data is passed to the receiving Transport service user in the same order in which it
was transmitted.
Congestion avoidance. Congestion occurs when a network, or part of a network, is
overloaded and has insufficient communications resources for the volume of traffic.

CHAPTER 6.

TRANSPORT LA YER

Congestion can re!3ujt in messages being discarded by the Network layer. Recovery
from lost messages is expensive, wasting system and network resources as well as
reducing the users' throughput and increasing the delay.
The Transport layer employs congestion avoidance algorithms to adjust the load such
that the network operates at a point where message loss due to congestion is unlikely.
It uses information provided by the Network layer and attempts to keep the users'
throughput to a maximum while sharing network resources fairly. Additionally, it
employs recovery mechanisms to respond to transient overloads which cause message
loss.

Segmentation and reassembly. Transport permits its users to transmit extremely large
messages. If the underlying ~etwork service is unable to accept a single message of
the desired size, Transport performs segmentation of the user message in to smaller
messages for transmission and the reconstruction of the complete message from the
received segments.
Multiplexing. When operating over a connection-mode Network service, Transport permits more than one Transport connection to be multiplexed onto a single Network
connection. Each Transport message carries a connection identifier, assigned during
connection establishment, which permits the receiving Transport to determine the
connection with which the message should be associated.
Concatenation. Transport may concatenate certain messages from the same Transport
connection and pass them to the Network layer as a single message for transmission.
This is sometimes referred to as piggybacking and reduces the number of Network
layer messages transmitted.
Error detection and recovery. The Transport Layer provides detection and recovery
from loss, duplication, corruption and misordering occuring in lower layers. It employs message numbering and a basic acknowledgment mechanism to ensure that
messages are delivered, and an optional checksum capability to detect message corruption.
Management. The Transport layer provides facilities to monitor and control the operation of the Transport layer itself and of individual Transport Connections.

6.2

Transport Layer Protocols

Two different protocols can be operated by the DNA Transport Layer. The first of these
is the OSI Transport protocol specified in International Standard ISO 8073. This standard
defines five classes of protocol, numbered 0 to 4. DNA provides classes 4, 2 and O. The
OSI Transport protocol is being introduced into DNA for the first time in Phase V. As
well as being available for communication between Phase V DNA systems, it also permits
Transport layer communication with non-DNA systems which implement this protocol.

6.3. OS! TRANSPORT PROTOCOL

The second protocol which may be operated by the DNA Transport layer is called the
Network Services Protocol (NSP). NSP was designed specifically for DNA and has been a
part of DNA since its inception. It offers communication between Phase V DNA systems
and provides backwards compatibility with Phase IV DNA systems.
The DNA Transport layer integrates these protocols and offers a single, consistent
service interface to the Transport service user. The choice of protocol for a particular
instance of communication is made during connection establishment. In practice, there
exist certain differences between the protocols which must be reflected in the Transport
service. Most such differences are minor and the Transport service user may use the
protocols in terchangeably.
The DNA Transport layer operates over the DNA Network layer and can use both types
of service provided by the Network layer: connectionless-mode (CLNS) and connectionmode (CONS). Certain of the Transport protocols may only be used with a particular type
of Network service. Table 6.1 shows the combinations of Transport protocol and Network
service which are supported.

NSP

OSI
Class 0

OSI
Class 2

OSI
Class 4

CLNSi

Yes

CONS

Yes

Table 6.1: Transport Protocols and Network Services

6.3

OS1 Transport Protocol

The OSI Transport protocol is defined by International Standard ISO 8073. This standard
defines five classes of protocol, the following three of which are provided by DNA.
class 0
class 2
class 4

Minimal features
Multiplexing class
Error detection and recovery class

These classes are actually distinct protocols which share a common message structure
and employ a common connection establishment procedure. During connection establishment the class of protocol to be operated is dynamically agreed by the participants.
The class 4 protocol offers the richest set of capabilities and can operate over both the
connectionless-mode and connection-mode styles of Network service. This protocol can do
everything that the other classes can do and is suitable for all purposes. It is expected to
be the most widely used.
Classes 2 and 0 are simpler protocols and have successively reduced capabilities. In
particular, they can only operate over the CONS. Use of the class 0 protocol is required

CHAPTER 6.

TRANSPORT LA YER

by the CCITT X.400 messaging standards when communicating with an AdministratIOn
Domain. Provision of the class 2 protocol offers a further choice which will increase the
level of interoperability between Phase V DNA systems and non-DNA systems.
The following sections describe in more detail some of the major features of the OSI
Transport protocols and how they are operated by DNA.

6.3.1

Connection Establishment

Connection establishment is an operation during which the class of protocol to be operated and other characteristics of the connection are agreed between the participants. A
connection is initiated by the transmission of a connection request message which identifies
the target Transport service user at the remote system. It also indicates a preferred class
of protocol and any acceptable alterna.tives, and various characteristics of the desired connection; for example, the maximum message size and whether subsequent messages should
carry checksums to guard against corr"..lption. The received connection request is accepted
by the transmission of a connection confirm message. This message carries the choice of
protocol to be used, and the actual characteristics for the connection negotiated according
to rules defined by ISO 8073.

6.3.2

Data Transfer

Once connection establishment has been successfully performed, the Transport Connection
enters the data transfer phase. The communicating service users are then provided with a
full duplex path for the exchange of data.

6.3.2.1

Flow Control

Classes 4 and 2 operate data message numbering for reliable, sequenced delivery. A sequence number from a circular number space is assigned to each data message transmitted.
As data messages arrive at their destination, the receiver returns an acknowledgment message to the transmitter indicating the message number up to which data messages have
been successfully received.
These protocols implement flow control according to the notion of credit. This means
that the receiver of data tells the transmitter how many messages it is prepared to receive
at a given point. Flow control operates in conjunction with the message numbering space.
The class 0 protocol has no explicit flow control. It does not number data messages
and does not send acknowledgment messages. Instead, flow control in the CONS is used
to match the speeds of transmitter and receiver.

6.3.2.2

Data Retransmission

The class 4 Transport protocol employs a retransmission timer to detect message loss. The
timer is started when the message is first transmitted and stopped when the message's
acknowledgment is received. If the timer expires, the message is assumed to have been lost
and is retransmitted.

6.3. OS! TRANSPORT PROTOCOL

The interval of the retransmission timer must be chosen carefully such that it is short
enough to ensure that loss of a data message is detected quickly, but not so short that
messages get retransmitted unnecessarily. DNA operates an adaptive algorithm which
maintains an average estimate of the round-trip delay on each Transport Connection. The
value used for the retransmission timer is then a simple function of the estimated roundtrip delay. The estimate is updated by continually sampling the round-trip delay and
incorporating new readings with the current estimate. Consequently, the value used for
the retransmission timer varies in sympathy with changes in the round-trip delay.

6.3.2.3

Congestion Avoidance

In DNA, operation of the class 4 protocol over the CLNS employs two algorithms which
adapt to the changing load on the network and prevent congestion.
The first operates in conjunction with the Network service and employs its congestion
experienced indication (Section 5.3.1). If, during a sample period, the number of received
messages which encountered congestion exceeds an architecturally defined threshold, the
flow control window size is reduced. This, in turn, reduces the number of messages which
may enter the network and so reduces the load. If the threshold is not exceeded, the
window size may be increased.
The second algorithm is invoked whenever message loss occurs on a connection. This
is assumed to indicate serious congestion and the transmitter then restricts the speed at
which it injects further data messages into the network. It operates a local credit window
which is smaller than the full window which it has been allocated by the receiver. This
local window is initially assigned a value of one, permitting the message which timed-out to
be retransmitted. Its value is then increased by one each time the number of data messages
for which acknowledgments have been received since the last change becomes greater than
the current value of the local credit window. This mode of operation is continued until
the local credit window grows to the value of the actual credit window assigned by the
receIver. Normal operation is then resumed.

6.3.2.4

Expedited Data

Protocol classes 4 and 2 provide an additional data transfer service called expedited data. l
This service permits a single, short data message to be transmitted which will bypass any
blockage due to normal data flow control. The class 0 protocol does not provide this service.

6.3.2.5

Reassignment After Failure

When the class 4 protocol is operated over the CONS, DNA attempts to maintain the
Transport Connection even if the underlying Network Connection fails. This is achieved
either by reassigning the Transport Connection to some other existing Network Connection,
or by establishing a new Network Connection for the purpose. The reassignment operation
is transparent to the Transport service user.
IDespite what its name may imply, expedited data is not necessarily transferred any faster than normal
data.

CHAPTER 6. TRANSPORT LAYER

Disconnection

6.3.3

A Transport Connection may be disconnected at any time, either by one of the Transport
service users, or by one of the peer Transport modules. In protocol classes 4 and 2, a
disconnect request message is transmitted and the receiver responds with a disconnect
confirm message. In protocol class 0, the Transport Connection is terminated simply by
disconnecting the underlying Network Connection.

NSP

6.4

The second Transport layer protocol provided by DNA is called the Network Services Protocol (NSP). This protocol has many similarities with the class 4 OSI Transport protocol
and provides an almost identical service. This section describes certain features of NSP
operation in more detail.

6.4.1

Connection Establishment

NSP establishes, maintains, and destroys Transport Connections by exchanging control
messages with a peer NSP module. An established connection is made up of two separate
data subchannels, each carrying messages in both directions:

Normal data subchannel. This subchannel carries Data-Segment messages between two
NSP modules.

Other-data subchannel. This subchannel carries Expedited Data messages and Flow
Control messages.

6.4.2

Data Transfer

6.4.2.1

Segmentation and Reassembly of Data

The Network layer limits the amount of user data that NSP can send in one message.
Taking normal data from its client's buffers, NSP breaks it up, if necessary, into segments.
Each segment is numbered and sent along with its number and other control information
in a Data-Segment message to the receiving NSP module. The receiving NSP module uses
the sequence numbers to reassemble the data segments in correct sequence in the receiving
user buffers. NSP segments normal data only. Expedited data messages do not require
segmenting since their limited size guarantees that they will always fit in a single Network
layer message.

6.4.2.2

Flow Control

NSP's flow control mechanisms ensure that data is not lost for lack of receiver buffering
capability and that deadlocks do not occur. Both normal and other-data subchannels are
flow-controlled. When a Transport Connection is formed, each NSP module informs the
other of the way it wants to control the flow of normal data as a data receiver. The
receiving NSP chooses either of two types of normal data flow control:

6.4. NSP

On/ Off only. The receiver tells the transmitter to stop or start sending data.
Segment with On/ Off. The receiver sends a request count of the number of segments
it can accept. In addition, the receiver can always tell the transmitter either to stop
sending data unconditionally or to start sending data under the normal request count
conditions.
The receiver also controls other-data flow with an Other-Data request count. The Data
Request message, Expedited Data message, Other-Data Request message, and the OtherData-Ack message may contain an acknowledgment for a Data-Segment message and/or
an Other-Data message. This ability, to combine acknowledgments with data and control
messages, reduces the number of NSP messages required per user message.

6.4.2.3

Data Retransmission

The NSP modules at each end of a Transport Connection positively acknowledge received
data. If the transmitting NSP fails to receive a positive acknowledgment during a timeout
interval, it retransmits the data. As for the class 4 OSI Transport protocol, the timeout
interval is adjusted dynamically based upon the round-trip delay observed by the transmitter.

6.4.2.4

Congestion Avoidance

While the flow control mechanisms protect against lack of receiver buffering, they do not
deal with insufficient resources in other parts of the network. NSP employs congestion
avoidance mechanisms, similar to those of the OSI Transport protocol module, to adapt
to the changing load on the network.
The primary difference between the NSP and OSI Transport congestion avoidance
mechanisms is that the congestion information reported by the Network layer is relayed to
the data-transmitting NSP. Based upon this congestion signal, the data-transmitting NSP
adjusts the maximum number of sent but unacknowledged data segments it is allowed to
have outstanding.

6.4.3

Disconnection

A connection may be disconnected at any time, either by one of the communicating Transport service users, or by one of the peer NSP modules.

CHAPTER 6.

TRANSPORT LAYER

Chapter 7

Session Control
The Session Control module, a client of the Transport layer, provides functions for systemdependent process-to-process communication, name to address mapping, and protocol selection. These functions bridge the gap between the Transport layer services and the
functions required by applications running under an operating system.

7.1

Functional Description

Session Control functions include:
• Matching incoming connect requests to the appropriate users of Session Control.
• Managing Transport Connections on behalf of users.
• Enforcing access control policies to restrict communication between users and between Session Control modules.
• Mapping from a DNA Naming Service object name to protocols and addresses.
• Selecting communications protocols supported by both the local and a remote end
system.
• Maintaining in the Namespace the protocol and address information corresponding to
local objects (see Chapter 8 on the Naming Service for the definition of a Namespace
and an object).

7.2

Session Control Functional Components

The Session Control module enhances the Transport service available to a network application. The purpose of Session Control is to form a bridge between the applications
requiring Transport service and the Transport module which actually creates, maintains,
and destroys Transport Connections.
55

CHAPTER I.

Session Con trol C ser In terf ace

SESSIOS GOSTROL

): aming Service
Clerk Interface

!
Address Selection

Address Resolution

DNA Naming Service Clerk

Connection Con trol

Transport user Interface

Figure 7.1: Session Control Components
Session Control is composed of three functional components, Connection Control, Address Resolution and Address Selection. The relationship among these functional components and between these components and the DNA Naming Service (see Chapter 8) is
shown in Figure 7.1. The vertical arrows indicate the flows of information with componen ts
below providing services to those above.

7.3

Connection Control

Connection Control is concerned with the system-dependent functions related to creating; maintaining, and destroying Transport Connections. Connection Control processes
requests for Transport Connections using the protocols and addresses passed by the application. It also enforces the locally defined access control policies.
The functions provided by Connection Control include:
• Requesting an outbound Transport Connection to an application based upon the

7.3. CONNECTION CONTROL

addresses and protocols specified in the request.
• Receiving an incoming Transport Connection request from the Transport layer.
• Validating access control information.
• Sending and receiving data.
• Monitoring a Transport Connection.
• Terminating a Transport Connection.
Many of these functions are similar to those that Connection Control requests from Transport. In response to requests from users, Connection Control makes corresponding requests
to Transport. The value added functions that Connection Control provides over the basic
Transport Connection service are described in the following sections.

7.3.1

Requesting a Connection by Destination Address

A connect request in which an application specifies all the protocols and addressing information is handled exclusively by the Connection Control component.
Connection Control performs the following functions:
• Formats connect data.
• Issues a connect request to the Transport layer protocol module specified by the
application.
• Starts an outgoing connection timer, if the user so requests. Expiration of the timer
prior to an acceptance or rejection from the Transport layer causes Session Control
to disconnect the Transport Connection for the source application.

7.3.2

Receiving a Connect Request

Upon receipt of an incoming connect request from the Transport layer, Connection Control
performs the following tasks:
• Parses connect data to obtain such information as source and destination application
descriptors and access control information.
• Validates any access control information.
• Identifies, creates, or activates a destination application context.
• Delivers the incoming connect request to the application.
• Starts an incoming timer when the connect request is delivered. Expiration of the
timer before the application accepts the connect request causes Session Control to
issue a reject to the Transport layer.

CHAPTER 7. SESSION CONTROL

7.3.2.1

'Validating Access Control Inforluatiull

Connection Control uses access control information included in an incoming connect request
to perform system-dependent validation functions. The information provided may be either
an explicit access control string (including a password) or a request to invoke a proxy on
beh alf of the requesting user.
Proxy mapping is a mechanism by which a user on one system in a network can be
given access to accounts on another system in a network without knowing the access control
information of the target accounts. This is accomplished by setting up a mapping on the
target system between the remote user and local, or proxy accounts. Upon receipt of a
connect request containing a (possibly null) proxy account name and an indication that a
proxy is desired, the appropriate proxy account is selected.

7.3.2.2

Identifying Applications

Connection Control executes a system-dependent algorithm to ascertain if an existing application corresponds to the destination application specified in an incoming connect request.
Connection Control may have an operating system interface to create a new user context
in which to run the application. Once the identity of the application is determined, the
connect request is passed to this context.

7.3.3

Sending and Receiving Data

This is a system-dependent function that handles application requests to send and receive
data. These functions are passed directly to the Transport layer.

7.3.4

Monitoring a Transport Connection

If requested by the user, Connection Control will:
• Detect probable network disconnection between the systems as reported by the Transport layer .
• Detect a failure to respond promptly to a connection request.

If either of these events is detected, then Session Control terminates the Transport Connection.

7.3.5

Disconnecting or Aborting a Transport Connection

If the application requests disconnection, Session Control will wait until all previously
transmitted data has been acknowledged by the remote transport module before issuing a
disconnect request to Transport.
Session Control immediately issues a disconnect request to the Transport layer in response to an application abort request. Previously transmitted but unacknowledged data
may not be delivered to the remote application.
Notification of a Transport layer disconnect or abort initiated by the remote Transport
is passed directly to the application along with any disconnect data.

7.4. ADDRESS RESOLUTION

7.4

Address Resolution

Address Resolution maps from a Naming Service object name to the identifiers of protocols
and corresponding addresses which are mutually supported by the local system and the
remote system(s) on which the named object resides. A Protocol Identifier is a character
string name for a protocol.
Accompanying these protocol identifiers is the address and protocol specific information necessary for protocol entities to exchange data. From this information, a sequence of
protocols and associated local and remote addresses may be selected to attempt comrnunication with the named object.
Previous versions of Session Control required each application to maintain, via network
management, a static node name to address mapping table. The size of the table, and
consequently the number of node names known to Session Control, was limited by local
storage considerations. Furthermore, the information in this mapping table was sometimes
out of date resulting in incorrectly addressed messages.
By relying on the Naming Service to store the mapping from names to addresses, the
number of destinations accessible to any given system is no longer constrained by the quantity of local storage available. Because the Address Resolution component automatically
maintains this name to address mapping in the Naming Service for local objects, the data
is more up to date.
The functions provided by Address Resolution include:
• Maintaining in the Namespace the protocol and address information corresponding
to local objects.
• Obtaining information about an object from the Naming Service for the purpose of
establishing the sequences of protocols and address information for communication
with the object.
• Caching protocol and address information to enhance performance.

7.4.1

Towers

In previous versions of DNA, there was no choice of the protocol operated at each layer
below the application, and each system had only a single Network layer address. In Phase V,
a system may support multiple Transport protocols (for example, NSP and OSI Transport)
and may have multiple Network layer addresses.
In order for applications to communicate, they must agree upon the protocols that
both will employ and the operational parameters of these protocols. In addition, these
users must possess address information that indicates to each layer of protocol where to
deliver data. This information is encoded in a Tower.
A Tower is a Protocol Sequence along with associated address and protocol specific
information. A Protocol Sequence is an ordered list of protocol identifiers. Associated with
each protocol identifier in a Tower is a component of the address and other protocol specific
information. The address information indicates the access point through which this layer

CHAPTER I.

SESSIO;'\T COSTROL

Addressing Inforrn (1 hon

Layer n -:- 1 protocol identifier

Layer 'n - 1 parameters and addressing
to reach n - 2 layer module

Layer n protocol identifier

Layer n parameters and add ressing to
reach n . . ;. . 1 layer mod ule

Layer n - 1 protocol identifier

Layer n - 1 parameters and addressing
to reach n layer module

Figure 7.2: Structure of a Tower

provides service to the next higher-layer protocol in the sequence. Other protocol-specific
information may be included in this field. Figure 7.2 illustrates the structure of a Tower.
Typically, a tower will extend from the DNA Application layer to the Network layer and
multiple Towers will be associated with an object.

7.4.2

Establishing Protocol Sequences for Communication

The Address Resolution component returns, in response to a client request, the protocol
sequences supported by both the local system and a remote system. A client of the Address
Resolution service passes both the name of a remote object and a local set of protocol
sequence for all layers above Session Control. The protocol sequences passed by the client
are combined with those below Session Control to determine the set of local protocol towers
of interest. The local protocol and address information supporting Session Control from
below is supplied by the Transport layer.
The Address Resolution component will then call the Naming Service Clerk to obtain
the set of Towers associated with the object name. The protocol sequences from the Towers
corresponding to the named object are compared with the local set of Towers to find the
intersection of matching sequences. A pair of protocol sequences are said to match if the
protocol identifiers in one member of the pair are identical to, are in the same order as, and
map one to one with those in the other member. Each matching sequence found is returned
to the requester along with the address information from both the local and remote Towers.
A user may then select from this set the protocols and addresses to attempt communication with the object.
To improve performance, protocol sequence and address pairs are cached for future use.
An application may request that the information cached about a name be discarded and
new protocol sequences and address pairs generated. Typically, this would be requested if
a connection attempt fails.

7.5. ADDRESS SELECTIOIV

7.4.3

Maintaining the Towers in the N amespace

Because the Network layer addresses are autoconfigured, the addresses associated with
protocols supporting Session Control may change with time. Session Control includes a
mechanism by which it can be asked to maintain the protocol and address information
(that is, the set of Towers) in the namespace for a set of object names.
Using information about the higher layers which is passed to Session Control from
above, and information about the lower layers requested by Session Control from each
underlying Transport protocol module, Session Control updates the protocol and address
information stored in the namespace for each such object by making requests to the Naming
Service Clerk.

7.5

Address Selection

Address Selection allows an application to initiate connections based upon the name of an
object. This relieves the user of both selecting protocols below Session Control and locating
address information. This component relies upon the services of Address Resolution to
perform name to protocol sequence and address mapping, as well as Connection Control
to establish Transport Connections.
Address Selection attempts to find a set of protocols mutually supported by the local
and remote systenls. A protocol sequence and associated address information are selected
from this set based upon local system policy. The selected protocols and address information are then used for communication with the remote object. In the face of failures,
Address Selection will retry Transport Connection establishment with another member
from the set of mutually supported protocols and address information.

7.5.1

Requesting a Transport Connection by Destination Name

An application may request a Transport Connection to a named object either by Node
Name or by any object name.

7.5.1.1

Connecting by Node Name

For Phase IV style naming, an application may present Session Control with the destination Node Name and application address information. For compatibility with existing
applications, Address Selection accepts a six-character node alias which it maps to the
Node Name. Through the services of the Address Resolution component, the transport
and network layer protocols and associated addressing information will be generated.

7.5.1.2

Connecting by Object Name

The user may present Session Control with an object name. Through the services of the
Address Resolution component, both the higher layer addressing information as well as the
transport and network layer protocols and addresses will be generated.

CHAPTER 7. SESSION COIVTROL

7.5.1.3

Oruel'ing the Protocol Sequences

As described above, the object name passed by an application in a connect request is
mapped to a set of protocol sequences and addresses via the Address Resolution services.
The Address Selection component then orders the members of this set in a system-specific
fashion. The services of the Communication Control component are subsequently invoked
to attempt Transport Connection establishment using the the first protocol sequence and
addresses from the ordered list.
7.5.1.4

Retry on Failure

If the initial attempt to establish a Transport Connection fails, the next protocol sequence
and associated address information are selected from the ordered set. Transport Connection establishment is again requested with these new parameters. The Address Selection
component will continue to step through this ordered set and retry connection establishment upon failure until either the set is exhausted or the reason for failure indicates that
further attempts would be futile.

Chapter 8

Naming Service
The Naming Service is a new component of DNA for Phase V. It is closely coupled with
the Session Control Layer, but its functions are directly available to network applications
as well. The fundamental purpose of the Naming Service is to enhance the logical organization of large networks by allowing the names of network applications, systems, and other
network-accessible objects to be independent of their physical location (address) in the network. Just as the Network layer provides the "glue" that organizes the physical structure
of the network from its topology of lines and systems, the Naming Service provides the
glue that organizes the logical structure of the network from the available application-level
resources.
The Naming Service associates strings of characters, known as primitive names, with
a set of attributes of the named object. For example, one of the named objects processed
by the Naming Service is the node object.! The name of each system in the network is
stored in the Naming Service. Along with the name an attribute is stored which contains
the system's network address. This allows applications to refer to systems by a consistent,
global name which is guaranteed to be unique. Storing system names in the Naming Service
also eliminates the need for a local node database, which performed a similar function in
DNA Phase IV.

8.1

Naming Service Concept's

Each name stored in the Naming Service refers to a single, unique object. Names for objects
are recorded in Directories, which themselves have primitive names. A directory contains
three types of entries, object entries, child directory entries, and soft links. An object
entry consists of the object's name and a set of attributes for the object, most prominent of
which is the Network Address where the object currently resides. Child Directory entries
link the directories together into a rooted tree, in which there is a single path from the
Root Directory through a set of child directories, to the desired named object. A soft link
is a form of "alias" or "indirect pointer" which allows a single entry to be reachable via
more than one name. Soft links both provide for user flexibility by allowing a namespace
ISee Section 10.1.1 for a description of how this is used by network managment.

64
to be viewed as a directed gratJli raLlIer than as a pure trfe. unci pro':icie a graceful '.'::-lY to
reorgamze a namespace.
A tree of directories; starting at a rooL is called a narnespace. ~vlultiple namespaccs may
coexist on the same network, but they are completely disjoint. 2 A namespace is uniquely
identified by a IV amespace Unique Identifier (NS UID). assigned when the namespace is
created. I\amespaces do not have names as far as the :\aming Sen'ice is concerned: but
are assigned a f'h'ckname so clients do not have to directly process the :\SCID.

8.2

The Semantics and Syntax of Names

8.2.1

The Semantics of Names

A Name is a complete path specification from the root directory of a namespace to the
particular entry in the directory of interest. Sometimes, however, it is necessary to refer
to the part of the name which specifies a particular object entry or child directory entry
in a directory as opposed to the entire Name. In these cases the terms simple name and
full name respectively are used. Every object entry and every soft link has a simple name.
Every directory has a simple name with the exception of the root directory of a namespace,
which has no name. A simple name denotes exactly one of either an object entry. a child
directory entry, or a soft link in a given directory.
All Names processed by the Naming Service through the client interface are full names,
which have the property of absoluteness or referential transparency. This means that a full
name refers to the same entry no matter which client provided the name, and that these
full names may be freely passed outside of the Naming Service (for example, on pieces of
paper) from one client to another without the possibility of confusion. Full names can be
quite cumbersome to use, however, since they can get rather long and contain the NSUID
of the namespace. While the Naming Service does not specify any mechanisms to alleviate
this problem, Digital operating systems on which the Naming Service runs contain some
native mechanisms for nicknames, which are a purely local shorthand for all, or part of a
full name (for example, VMS logical naInes).

8.2.2

The Syntax of Names

There are two syntaxes for Naming Service names, the external name and the internal
name. The external name is the syntax used for the human-readable form of a name.
The internal name is the syntax of the name as passed across the client interface with the
Naming Service. The external name is designed for readability while the internal name is
designed to be convenient to encode in programs, protocols, and databases.
An external name consists of two parts: a namespace nickname and a sequence of
simple names. The namespace nickname is translated locally by the Kaming Service into
the equivalent NSUID. If a namespace nickname is omitted from an external name, the
2In general it is undesirable to have multiple name:::paces on a net.work because they are significantly
harder to use and manage than a single namespace. The existence of multiple namespaces i::: a temporary
phenomenon which arises when previously disjoint networks are merged into a single network.

8.3. CONTENTS OF NAMESPACE ENTRIES

Naming Service chooses a default namespace based on a local, implementation-dependent
algorithm.
The namespace nickname and each simple narne in the sequence that represents the
full name is an ordered list of strings of letters, digits and certain punctuation characters
from the Digital multinational character set. The case of each string is preserved by the
Naming Service: a name registered with a mix of upper and lower case characters will
appear when later retrieved ~xactly as entered. Lookups however, are case-insensitive.
The strings "AbCoE", "ABCOE" and "abcoe" all represent the same simple name.
The simple names in the sequence are delimited by a period ".". If a client desires to
use the "." (or any other disallowed punctuation character) as part of a simple name the
name must be enclosed in double-quotes. Figure 8.1 contains some example full names.

Parts. widgets.left-handed .SMOKESH IFTER
Government:Treasury. Bills .CurrentSeries
DEC:. E ngi neering. Networks.Arch .Specs
U LTRIX.Sources. "OSITransport.c"

Figure 8.1: Examples of External Names

8.3

Contents of Namespace Entries

All Naming Service entries (objects, directories, son, links, and child directories) consist of
a set of attributes and their associated value or vakes. Attributes are of two types. When
an attribute may take on more than one value, it is called an attribute set; if it may have
only one value at a time it is called a single valued attribute. An attribute set operates as
a "dictionary" which allows the following operations:

Redundant Insert: A new value may be inserted in the set. If the new value matches a
value already present, a separate entry is not made. The client has no control over
the ordering of the members of the set.

Redundant Delete: An existing value may be removed from the set. If the value is not
present, no error is returned.

Full Lookup: The entire contents of the set may be returned
Attributes fall into two general categories, globai attributes and class specific attributes.
Global attributes are those whose meaning is the same for all entries stored by the Naming
Service. Class specific attributes are attributes whose definition is dependent on the value
of one of the global attributes: the class attribute.

CHAPTER 8.

8.3.1

IVA1\lING SERVICE

<;}obal Attributes

Table 8.1 lists some of the global attributes defined by the r\ aming Service along with a
description of the function of the attribute and an indication of the kind of i\aming Service
entry with which it is used.

8.3.2

Predefined Object Classes

Object entries, in addition to their use by clients directly, are also used by the N"aming
Service itself to provide certain functions, such as the ability to give clearinghouses humansensible names, and to provide the notion of a group. Table 8.2 lists some object classes
that are predefined by the Naming Service.

8.4

Operational Concepts and Terminology

A namespace is stored in a partitioned, partially replicated database. The database is
partitioned because parts of the namespace are stored in different locations (that is, on
differen t systems). The database is partially replicated becallse part of the namespace
may be simultaneously stored in multiple locations. The unit of both partitioning and
replication is the directory. A collection of (copies of) directories stored on a particular
system is called a Clearinghouse. The partitioning is accomplished by controlling which
directories are stored in which clearinghouses. The replication is accomplished by storing
a directory in more than one clearinghouse.
Clearinghouses have names so that clients may conveniently refer to them, and so that
the Naming Service can find a clearinghouse by looking up its name in the namespace.
The naming of clearinghouses follows a set of strict rules which ensure that a name lookup
cannot fail because the clearinghouse in which the directory storing a particular entry
cannot be found.
Clearinghouses are either "active" or "inactive". When a clearinghouse is active at a
given system, that system is acting as a nameserver. A nameserver may be controlling
more than one clearinghouse simultaneously, especially when a clearinghouse has been
moved after the failure of a nameserver.
A copy of a directory stored in a particular clearinghouse is called a replica. In order
to simplify the algorithms for general namespace maintenance, one of the replicas of a
directory is designated to be the master replica for that directory. Master replicas perform
certain overhead functions that would be redundant if carried out independently by every
replica of a directory. A second kind of replica is the secondary replica. Creation of new
objects, directories, and soft links may be done at either a master or secondary replica.
Updates to existing entries and deletion of entries may also be done at either kind of replica.
A third kind of replica, the read-only replica, responds only to lookup requests and is not
permitted to perform creations, updates, or deletions on behalf of clients.
The Naming Service maintains a distributed database on behalf of its clients. This
database does not have the usual characteristics of a distributed database since it provides

8.4. OPERATIONAL CONCEPTS AND TERMINOLOGY

Attribute
UID

Entries
All

Description
A Unique Identifier for the entry. This is a space
and time unique identifier for the entry which
is assigned when the entry is made and is never
changed.

UTS

All

An Update Timestamp indicating the Timestamp
of the most recent update to an entry.

ACS

All

The Access Control Set for the entry. Each value
is an Authorization Entry (AE).

Class

Object

An Attribute used to classify object entries according to the type of object being named. This
attribute allows clients to correctly interpret the
contents of the class specific attributes of an entry.

Address

Object

Contains a set of Network Addresses where the
object currently resides.

Replicas

Directory

Each member of this attribute set identifies one
of the clearinghouses which stores a replica of this
directory.

AllUpTo

Directory

This attribute contains a bound on how out of
date various replicas of this directory are. All
replicas of a directory are guaranteed to have received all updates whose timestamps are less than
the value of AllUpTo.

Convergence

Directory

ParentPointer

Directory

This attributes specifies how persistent a directory should be in trying to keep its replicas up to
date.
This attribute links a child directory to its parent,
allowing the Naming Service to workup the tree
as well as down. It is used to ensure that parent
and child directories always point to each other
during normal operation, and is used during directory creation to allow a child directory to link
itself into the tree by creating a child directory
entry in the parent directory.

LinkTarget

Soft Link

Contains the full name of the entry the soft link
points to.

Table 8.1: Naming Service Global Attributes

CHAPTER 8. NAMING SERVICE

Object Class
Group

Description
Defines a group. The members of the group may be
either principals (names which identify clients of the
name service) or other objects of class Group. A
client interface operation: TestGroup, is provided
to test for Group Membership. This operation is capable of recursively examining groups within groups,
and can detect loops in group membership.

Clearinghouse

The Clearircghouse Object is used internally by the
Naming Service for locating Clearinghouses.

Table 8.2: Naming Service Predefined Object Classes

very "loose" consistency guarantees in order to allow high levels of partitioning and replication. A client may get different anSWLfS depending which replica of a directory is queried
if updates are still being propagated through the network.
Updates to a namespace are timestamped and applied such that the update with the
latest timestamp wins. The updating algorithms are designed such that all updates are
idempotent (multiple applications of an update to the database have the same effect as a
single application of the update) and commutative (updates may be applied in any order
with identical results).
The primary algorithm for producing convergence among the replicas of a directory is
called the Skulker. Skulks operate independently on each directory in a namespace. Successful completion of a skulk operation ensures that all replicas of a directory have absorbed
all updates applied to any replica prior to the time the skulk started. The more frequently
skulks are run, the more up to date all replicas of a directory are maintained. Skulks
are expensive, however, so there is a tradeoff between resources (bandwidth, processing,
memory) and timeliness. Clients can control the frequency of skulks either by adjusting
an attribute of a directory, or initiating skulks themselves.
A typical skulk operates as described below. The master replica and all secondary
replicas of a directory are linked together into a virtual ring, using the RingPointer attribute
of each replica. The ring keeps multiple skulks of a single directory from getting in each
other's way. Each skulk of a directory does the following:

• Gathers up all updates made (to the master replica and any secondary replicas) since
the last skulk and applies them to the clearinghouse where the skulk is running,
• Spreads all the gathered updates to all replicas of the directory, and
• Informs all replicas of the timestamp of the latest update all of them are guaranteed
to have seen.

B.S.

FUNCTIO~NAL DECOlv!POSITION OF THE NAMING SERVICE

8.5

Functional Decomposition of the Naming Service

The Naming Service is functionally divided into a number of major modules. Figure 8.2
illustrates the relationships among the modules of the Naming Service and the protocols
employed for communication among the modules.

Clerk

S.Protocol

C.Protocol

M.Protocol
Transaction
Agent

Transaction
Agent
C.Protocol

Update
Sender

P.Protocol
N ameserver 1

Update
Listener
N ameserver 2

Legend:
S.Protocol
C.Protocol
M.Protocol
P.Protocol

Solicitation and Advertisement Protocol
Clerk-server Protocol
Directory Maintenance Protocol
Update Propagation Protocol

Figure 8.2: Naming Service Modules and Protocols

A Naming Service client accesses the Naming Service through a client interface. A
summary of the functions available to clients of the Naming Service is provided in Table 8.3.

8.5.1

Clerks

The client interface is provided by a Naming Service Module called a clerk. The clerk is
the only module of the Naming Service that must reside in all systems; other modules
reside only in nameservers. The clerk ascertains the appropriate nameserver to process a
request, invokes the Clerk-Server Protocol (C.Protocol) to speak to as many Nameservers as
necessary to satisfy the request, and contains a cache of recently accessed entries to improve
performance. The relationship of the clerk to a nameserver is illustrated in Figure 8.2.
Clerks are responsible for learning about at least one nameserver that can either pro-

CHAPTER 8. NA:VIJ.NG SERVICE

Function

Description

EnumerateAttributes

Enumerates the attributes of an object entry~ directory entry, soft link, or clearinghouse

ReadAttribute

Returns the value(s) of the specified attribute

Modify Attribute

Modifies (or deletes) an attribute or attribute value

TestAttribute

Tests an attribute for whether a specified value is one
of its current values

CreateObject

Adds an object entry to the namespace

EnumerateObjects

Returns the names of object entries in a directory

DeleteObject

Removes an object entry from the namespace

CreateDirectory

Creates a child directory of the specified parent directory in the namespace

AddReplica

Adds a Clearinghouse to the replication list of a directory

Remove Replica

Removes a clearinghouse from the replica set of a
directory

DeleteDirectory

Removes the specified directory from the namespace

En umerateChildren

Returns information about child directories of the
specified parent directory

Skulk

Skulks a directory to force convergence of its replicas

CreateLink

Creates a soft link for an entry

DeleteLink

Deletes a soft link for an entry

EnumerateLinks

Enumerates the Soft Link entries in a directory

ResolveName

Follows a chain of soft links and returns the full name
of the entry pointed to. Cycles are detected.

TestGroup

Tests for group membership, allowing for recursively
defined groups and detecting cycles

Table 8.3: Summary of Naming Service Client Interface

B.S.

FUNCTIONAL DECOMPOSITION OF THE NAMING SERVICE

cess a request or provide information about other nameservers that may be useful. This
initial information is obtained via the Solicitation and Advertisement Protocol {S.Protocol}.
N ameservers periodically advertise their availability.

8.5.2

Nameservers

Nameservers consist of four modules. The Nameserver Control module provides the Management Interface and coordinates the overall operation of a nameserver - turning it on
and off, bringing clearinghouses online and offline, etc. This module is also responsible for
advertising the availability of the nameserver to any clerks that may be listening.
The Transaction Agent module processes the nameserver side of the Clerk-Server Protocol, performing operations requested by clerks. It accesses clearinghouses, performing
creations, updates, deletions, and lookups as specified by the clerk. It causes modifications
for an entry to be propagated to all replicas of the directory when a change is made. 3 The
Transaction Agent is also responsible for coordinating the creation: deletion, and modification of directories using the Directory Maintenance Protocol {M.Protocol}, communicating
with other Transaction Agents as necessary.
The Update Sender module is responsible for propagating any changes made by a Transaction Agent to all clearinghouses that need to know about the change. It uses the UpdatePropagation Protocol {P.Protocol} to accomplish this. The Update Listener Module receives
updates sent by the Update Sender and records the changes in the appropriate clearinghouse. The Update Sender and Update Listener are also responsible for ensuring that
clearinghouses can find each other when a new clearinghouse is created or if a clearinghouse moves. This is accomplished by clearinghouses modifying their own registration
information in the namespace and by maintaining appropriate pointers such that the root
of the namespace may be found by starting at any clearinghouse in the namespace.

3The actual propagation is carried out by the Update Sender module.

CHAPTER 8.

NAl'\JHNG SERVICE

Chapter 9

Support of X.25
Public data networks offering the X.25 interface are available in many countries and represent an economical alternative to leased and dial-up lines for many applications. CCITT
Recommendation X.25 defines a standard interface between a system, termed the Data Terminal Equipment (DTE) and another system, termed the Data Communication Equipment
(DCE).
The DCE is an access point to a Packet Switched Data Network (PSDN) implementing
the X.25 interface. The X.25 interface is structured into three levels:
• Levell, the Physical level, defines the characteristics of the physical link between the
DTE and DCE and resides in the Physical layer of DNA. The X.25 Recommendation
defines Levell only by reference to the X.21 standard for digital networks and an
interim X.21bis standard for access via analogue lines; X.21 is discussed in Chapter 2.
• Level 2, the Frame level, defines the mechanisms for reliable communication between
the DTE and the DCE. This level resides in the Data Link layer of DN A. The LAPB
protocol! is used at Level 2.
• Level 3, the Packet level, defines the format and meaning of the data field contained
in each frame. The remainder of this Chapter describes Level 3 facilities and the
functions they provide in DNA.

9.1

X.25 Service Functions

Levels 1 and 2 must be operational before packets can be sent on an X.25 virtual circuit
(Virtual Circuits are defined in 9.3.1.2) between a DTE and DCE. Level 3 functions are defined in terms of a set of procedures, formats and optional facilities supported at the packet
level user interface of a DTE (operating in conformance with the X.25 Recommendation).
These include functions for the establishment and release of circuits) the management of
circuit characteristics and packet flow-control. The Level 3 functions that are specified in
Phase V reflect the utilization of the X.25 packet level protocol in Phase V:
lLAPB is a subset of the High-level Data Link Control (HDLC) protocol.

CHAPTER 9. SUPPORT OF X.25

• To provide X.25 packet level communication facilities to user-written programs that
reside in the D?\A Application layer. These facilities allow a user program in a D~A
system to communicate with both DNA and non-DNA systems over an X.25 PSD):.
• To provide the Transport layer with a set of capabilities to support the Connection]tv! ode Network Service (CONS) over an X.25 PSDN by mapping between the abstract primitives and parameters of CONS, and X.25 procedures and packet types.
The CONS abstract primitives include those for network connection establishment,
network connection release, data transfer and reset.
• To provide the Network layer with basic, underlying capabilities to support the
ISO 8473 Connectionless-Mode Network Service (CLNS) over an X.25 subnetwork.
The Network layer manipulates X.25 service functions such that X.25 virtual circuits 2
can be used to realize routing over the subnetwork.

9.2

X.25 Modules

The following modules cooperate to provide the X.25 functions in Phase V DNA.
X.25 Gateway Access Module. This presents an interface to user-written programs on a
host DNA system and communicates with a Gateway system (a DNA system that is
a DTE on an X.25 PSDN). It resides in the Application layer of the host system.
X.25 Server Module. This resides in the DNA Application layer and presents local userwritten programs in the system with an X.25 access interface. When the system is
also an X.25 gateway it communicates with an X.25 Gateway Access Module over a
DNA Session Control connection.
X.25 Packet Level Module. This provides the functions of the X.25 Level 3 packet level
protocol. It is used by the X.25 Server Module to gain access to an X.25 PSDN.
LAPB Module. This provides the functions of the X.25 Level2 frame level protocol. It
resides in the Data Link layer of the system and is used by the X.25 Packet Level
Module to achieve reliable, frame level communication with the DCE.

9.3

X.25 Packet Level

Functional descriptions of Levell and Level2 of X.25 can be found in chapters on the DN A
Physical layer (Chapter 2) and Data Link layer (Chapter 3) respectively. This section
describes the features of Level 3, the packet level of X.25.
The use of X.25 circuits for ISO CONS and CLNS/Routing involves mapping from the
abstract requirements of the respective "service" to X.25 circuit characteristics for both
outgoing and incoming calls. Some or all of these abstract requirements can be codified
2Two DNA systems may communicate through an X.25 PSDN over a dynamically established data link

(DED) (Section 5.2.4) and use standard DNA protocols and utilities.

9.3. X.25 PACKET LEVEL

into a call parameter profile or template (see Section 1.3.1). The user interface to the X.25
packet level in Phase V uses the template as a mechanism that allows the in vocation of
default characteristics for outgoing calls and discrimination criteria for incoming calls on
X.25 circuits. Service specific templates are held in appropriate databases and maintained
by network management.

9.3.1

Basic Features of the Packet Level

9.3.1.1

Logical Channels

X.25 circuits are defined in terms of logical channels at Level 3. Each logical channel has
a unique identifier in the packet header. The range of channels which can be used by the
DTE is specified by the network administrator.

9.3.1.2

Virtual Circuits

The packet level establishes a virtual circuit in response to a user request. Logical channels
are used to offer permanent virtual circuit (PVC) and switched virtual circuit (SVC) user
facilities.
A PVC is a permanent association between two DTEs. A DTE can send and receive
packets to and from the associated DTE at any time. A logical channel is permanently
assigned to each permissible PVC. The number of permissible PVCs is agreed with the
PSDN administrator.
An SVC is a temporary association between two DTEs. It is initiated by a DTE sending
the network a call request packet containing the destination address. A logical channel is
allocated to the SVC at this time. On receipt of the call request packet the PSDN will send
the called DTE an incoming call packet on a free logical channel. If the call is accepted
the virtual circuit enters the data phase and data packets can be transferred.

9.3.1.3

Flow Control and Packet Sizes

The PSDN uses flow control in order to control the rate at which it accepts packets.
X.25 has independent flow control for each direction of transfer on a logical channel,
implemented at both the DTE and the DCE by employing a window mechanism. The
window size represents the maximum number of sequentially numbered data packets that
may be outstanding at any given time.
The More Data bit (M- bit) indicates to the destination DTE the boundaries of the
"original" data packet. A larger packet may be sent from the DTE (or the DC E) as
a sequence of consecutive, maximum length packets with the More Data bit set to on
followed by the last packet, with the More Data bit off. As the maximum data packet
size can be specified independently at each end of a circuit, their difference can cause
the number of packets arriving at the destination to be greater or less than the number
originally sent.
The window and packet sizes can remain fixed as agreed at subscription time or an optional facility called the Flow Control Parameter Negotiation allows negotiation of window
and packet sizes on a per virtual circuit basis.

9.3.1.4

CHAPTER 9.

SL'PPORT OF X.25

Interrupting

The normal packet sequence can be bypassed by use of interrupt packets. Interrupt packets
are non-sequenced; carry up to 32 octets of user data and can be delivered even when the
destination DTE is not accepting normal data packets.
The initiating DTE receives an interrupt confirmation packet for every interrupt packet
it sends; implying that the interrupt has been received and acknowledged at the destination
DTE/DCE interface. Each interrupt packet must be confirmed before another is sent.
9.3.1.5

Resetting

The reset facility is used to reinitialize a PVC or an SVC in the data phase. It removes all
data and interrupt packets, (for both directions of that circuit), that may be in the PSDN.
Reset can be initiated by either DTE or by the PSDN by sending a reset request packet.
The initiating DTE or DCE receives a reset confirmation packet to indicate resetting has
completed.
Resetting can cause the loss of packets that have been previously acknowledged at
the local interface. Therefore the user of Level 3 has to participate in the recovery and
re-synchronization following a reset.
9.3.1.6

Clearing

A called DTE has the choice of accepting or rejecting the call; the latter by using a clear
request packet. (The calling DTE will then receive a response indicating whether or not
the called DTE has accepted the call.)
Either DTE may clear an established call by sending a clear request packet. In response
the DCE sends a clear indication packet to the remote DTE. The remote DTE replies with
a DTE clear confirmation packet to the DCE which then sends a DCE clear confirmation
packet to the DTE that initiated the clear. The logical channel at either end is deassigned
after the sending or receipt of the respective clear confirmation packet.
9.3.1. 7

Restarting

A restart is used to clear all virtual circuits at the DTE/DCE interface. A restart indication
packet from the DCE forces all virtual circuits at a DTE to be cleared. A DTE sends a
restart indication packet to the DCE as part of its initialization procedure.

9.3.2

Optional Facilities of the Packet Level

The X.25 Recommendation defines a number of Optional User Facilities for use between a
DTE and DCE, enhancing the features available to user programs. Some of these facilities
are briefly described below.
The X.25 Recommendation categorizes some optional user facilities as being applicable
on a per virtual call basis. Others have to be agreed with the PSDN administrator and
therefore remain in force for the duration of the agreement with network management.

9.4.

X.2S GATEWAY ACCESS

9.3.2.1

Closed User Groups

The CUG facility enables the formation of groups of DTEs. Access to a DTE is limited to
members of the groups to which the DTE belongs. A CUG allows various combinations of
accessibility of member DTEs, including; outgoing access allowed, incoming access allowed,
outgoing calls barred, incoming calls barred and selection of a CUG on a per call basis.
A special case called the bilateral CUG enables pairs of DTEs to access each other while
limiting access to or from other DTEs that have not made bilateral associations.

9.3.2.2

Call Redirection

This facility allows redirection of calls destined to a DTE that is out of order, busy or
that has an explicit request set up to redirect all calls. Additional capabilities includes
specifying a list of alternative DTEs to try, and specifying a logical chain of DTEs for
continued redirection.

9.3.2.3

Network User Identification

This facility allows the DTE to provide information to the PSDN for security, network
management or billing purposes on a per call basis. The DCE can clear a call on the basis
of unsatisfactory identification.

9.3.2.4

Call Charging

A DTE which subscribes to Reverse Charging Acceptance agrees to accept incoming calls
which request reverse charging. A DTE subscribing to Local Charging Prevention optional
facility prevents the establishment of calls which the user must pay for. Charging Information is a facility that allows the DCE to indicate to the "DTE to be charged" , information
that will allow it to calculate the charge. It can be applied on a per call basis or can be
subscribed to apply to all calls for which the DTE will be charged.

9.4

X.25 Gateway Access

The X.25 Gateway Access facility enables programs in any host DNA system, containing
a Gateway Access Module, to gain access to an X.25 PSDN. It also enables incoming calls
to be directed to a user process residing at the host DNA system.
Gateway operation is accomplished by establishing a connection between the Gateway
Access Module and a DNA system that is a DTE on an X.25 PSDN. Such a DNA system
is termed the Gateway system. Programs using an X.25 gateway can access all the facilities
available at the DTE. Figure 9.1 illustrates the X.25 Gateway Access facility.
The communication between the Gateway Access Module and the Gateway system
is over a DNA Session Control connection and is based on the Gateway A ccess Protocol
(GAP) .

CHAPTER 9.

SUPPORT OF X.25

HOST SYSTEM
User Layer:
Accessing program:
User-written or
Digital-supplied

GATEWAY SYSTEM

DNA Application:

REMOTE SYSTEM (DTE)

X.25 Gateway Access
Module

X.25 Server
Module

User Layer:

Session Control
Transport
Network
Data Link
Physical

Cooperating program:

X.25 Level 3
X.25 Level 2
Physical

X.25 PSDN

DNA Network

Legend:
D TE - Data Terminal Equipment
DeE - Data Circuit Equipment

X.25 Leve13
X.25 Leve12
Physical

--Z--................

A physical communication link
A path through a network

Figure 9.1: The X.25 Gateway Access Facility

9.4. X.25 GATEWAY ACCESS

The Gateway Access Protocol defines messages that accomplish remote access of the
X.25 Level 3 service interface functions. Messages are sent and received on the Session
Control connection between the host system and the Gateway system.
The GAP message-types reflect functions available at the Level 3 interface and include
those for:
• Opening a PVC
• Making outgoing calls
• Accepting and rejecting incoming calls
• Clearing calls
• Resetting calls
• Transferring data
• Sending and receiving interrupt data
• Indicating failure conditions on a virtual circuit

9.4.1

X.25 Gateway Access Operation

A GAP dialogue between an X.25 Gateway Access Module and a Gateway system over a
Session Control connection begins with a user-written program opening a port (by placing
an Open Port call to the Gateway Access Module). It ends with the closing of the port by
the program.
Service calls by a user-written program to the X.25 Gateway Access Module are translated into GAP messages and sent to the Gateway system. Similarly, events at the packet
level interface of the Gateway system are translated into GAP messages to be received by
the Gateway Access Module. Three processes are involved in the X.25 Gateway operation:
• The User process, in a host DNA system, that accesses X.25 Level 3 functions by
issuing calls to the X.25 Gateway Access Module (resident in the same system).
• The X.25 Gateway Server process, in the Gateway system, that exchanges GAP
messages with the X.25 Gateway Access Module (over the Session Control connection
with the "host" system).
• The Remote DTE process on the remote X.25 system with which the Gateway system
communicates over an X.25 virtual circuit.
When a port is opened by the User process, information in an associated template
specifies whether the User process requires a PVC, wishes to make an outgoing call or
wishes to receive incoming calls. Opening a port causes a Session Control connection to
be established and the appropriate GAP message to be sent to the Gateway system.
Subsequent GAP messages depend on what was invoked. For example, a successful
outgoing call results in a message that indicates confirmation of the remote system having

CHAPTER 9.

SUPPORT OF X.25

accepted the call. Data transfer can then proceed between the User process and the Remote
DTE process on the remote system (as can other operations that are valid while in the
X.25 data phase).
Either the Cser process or the Remote DTE process can initiate shut down of the
association. Failures at the PSD::\ can also result in loss of connection between the two
processes. The Gateway Access !\1odule reports such events to the Cser process.

Chapter 10

Network Management
DNA network management allows system or network managers to control and monitor the
operation of a DECnet network. It allows network operating parameters to be configured
where configuration is necessary, or for fine tuning of these parameters following autoconfiguration. It allows the manager to startup and shutdown network components as
needed. It allows the network to be monitored, providing information related to network
traffic and performance. This data, collected in real time, can be reduced offline to produce
statistical and auditing information. It may also playa role in long term network planning.
DNA network management can provide information warning network managers of faulty
or failing network components, both hardware and software. The facilities provided allow
network problems to be detected. Following detection of a problem, a faulty component
can be isolated from the network, returning to service once repaired.
DNA network management is developed within the framework of a distributed system
management model. This design approach provides the following benefits:
• Network management is consistent across the architecture, which in turn allows products developed by different sources to be managed in a consistent way.
• Network management is modular, which allows systems to be as simple or as complex
as appropriate to the services they provide their users.
• Network management is extensible, which allows new functions to be added to DNA
and managed consistently with existing functions.
DNA fully defines the network management capabilities of each architectural layer or module.
Management of a DECnet network can either be distributed or centralized. Where it is
distributed, the extent to which it is distributed can be controlled. A DECnet network can
be managed by a human operator at a terminal, or by specialized network management
programs. DNA defines a set of primitive management functions which can then be combined and tailored to meet the specific requirements of a particular network configuration
or administration.
DNA network management is designed to utilize other DNA services using standard
interfaces, for example the DNA Naming Service or the Session Control layer. In general,
81

CHAPTER 10.

j\"ET\VORK 1\1A;YAGEA1EST

management functions are performed in the upper layers of D:\A. The services of the
Data Link layer are used directly to allow basic management operations such as loading
and dumping when not all network protocols are operating~ for example during system
initialization or failure.
Dr\A Phase V is based upon the emerging international standards for network management. However, the current draft proposals do not cover all aspects of managing networks.
In cases where standards are not currently being developed, D;\;A uses proprietary solutions with the intention to migrate towards international standards when these become
available.

10.1

DNA Network Management Model

DNA network management is designed within the framework of a distributed system management model. The model defines in a general way the management components in a
distributed system and the interfaces between them. There are two major components: directors and entities. Directors are defined as the management software used by a network
manager. Entities are the components that make up the network and which are managed.
Figure 10.1 illustrates the model.
Network
Manager

Director

Management
Protocol

Servic e interface
A
g
e
n
t

Service
Element

Entity

Clien t interface

Figure 10.1: DNA Network Management Model - Functional Components
DKA network management is used through a director. Directors, if they are to be used
by human operators, provide a user interface for this purpose. They interpret commands
entered at a management console, send these commands to a target entity for processing
and present the results back to the console. An entity has two logical parts: a service
element and an agent. The service element performs the primary functions of the entity, for

10.1. DNA NETWORK MANAGEMENT MODEL

example, provision of a layer protocol, while the agent provides the management interface
to the entity.
All directors use common mechanisms to communicate with the entities they manage.
The same director can manage local entities, and remote entities on many different systems.
If the managed entity resides on the local system, the director may use a local interface to
access it. If it resides on a remote system, a management protocol is used.
The management information and operations that pass between directors and entities
are described below:

Directives. Directives are the management commands issued by a director to an entity.
They allow a director to read and alter an entity's management attributes, or to
request it to perform a specific action.
Attributes. An attribute is a piece of management information maintained by an entity.
Each attribute has a name allowing it to be accessed by management. There are a
number of different attribute types, listed below:
• Identification.

An identification attribute identifies an entity to network man-

agement.
Characteristic attributes allow a manager to control the operating parameters of an entity. For example, the parameters which determine
the DDCMP polling rate, or the cost of a Routing circuit. In general, characteristics take default values when the entity is created, but these values may
subsequently be changed only by network management commands.

• Characteristics.

Status attributes allow a manager to inspect the current state of an
entity. Unlike characteristic attributes, status attributes may change without
management intervention, for example, in a protocol entity to reflect the changing state of a connection.

• Status.

Counters are attributes whose value indicates the number of times
an operation has been performed by an entity, or a particular condition has been
detected.

• Counters.

Events. An event is an occurrence of a specific normal or abnormal condition. For example, the discarding of a data NPDU by the Network layer because the destination is
unreachable.

10.1.1

Entity Hierarchy and Naming

A DECnet network is constructed out of many manageable components - the more systems
in the network, the more manageable components there are. To allow management of large
networks, the components must be organized into a logical structure, and named in a way
that allows a manager to deal with the complexity. The organization defined in DNA is a
hierarchical one with parent entities "above" their children in the hierarchy.
Each computer system in a DECnet network is a top-level component which is represented to network management by the node entity. A node entity is assigned a name which

CHAPTER 10.

NET~VORK MANAGEl'vfEI\'T

is unique throughout the network. This forms the basis for asslgmng names to "lesser';
entities in the system which will themselves have network-wide uniqueness. :\ode names
are registered with the DNA Naming Service together with an address by which the system
can be reached.
Below the node entity are module entities. Modules comprise a group of network
functions which together provide a particular service. For example, DN A defines Routing.
OS1 Transport and DDCMP modules, and many others. There is only one occurrence, or
instance of a module entity in a node. For this reason, modules can be uniquely identified
within a node by their class name, for example "DDCMP", and no further qualification
by instance name is required.
At the next level down in the hierarchy are subordinate entities, or sub-entities. These
are defined to allow management of some part of a module's functions. The DDCMP
module, for example, maintains Link sub-entities for each communications link over which
the protocol operates. "Link" is the class name of the sub-entity. Each instance of the Link
sub-entity requires further identification, to allow it to be distinguished from the others.
Further levels of sub-entity can be defined for each entity.
The entity hierarchy is used to form an entity name. An entity name comprises a global
part and a local part. The global part consists of the node entity name which identifies
the node in the network. The local part identifies the entity within the node. The local
name is a derived from the class name and instance name (if applicable) of the entity itself
and those of its parent entities, up to the level of the node entity. This is similar to a fully
specified mailing address, where the hierarchy might consist of the country, state, city,
street, house and individual.
A directive for an entity in a remote system is sent to the node agent of the system
and includes the entity name. The node agent uses the local part of the name to identify
a sub-entity to handle the directive. This process may be repeated by sub-entities on any
remaining portion of the name until the target agent is reached.
The entity hierarchy is defined to allow network components be uniquely identified for
management purposes. It imposes no restrictions on the actual management organization
of a DECnet network.

10.2

Network Management Operation

This section gives an overview of the major Phase V network management components and
operation.

10.2.1

Common Management Information Protocol (CMIP)

The DNA Common Management Information Protocol, or CMIP, provides an encoding of
the network management operations which can be performed on entities, and their parameters. DNA CM1P is based closely on the emerging ISO CMIP standard, currently at a
draft stage. It is in fact the combination of two protocols - the Management Information
Control and Exchange (MICE) protocol, used between a director and a node agent, and the
Management Event Notification (MEN) protocol. DNA CMIP is a simple Application layer

10.2. NETWORK MANAGEMENT OPERATION

request-response protocol which operates over DNA Session Control. It replaces the Network Information and Control Exchange (NICE) protocol of DNA Phase IV, and provides
operations to get and set management attributes, to request execution of management
actions and to report events.

10.2.2

Event Logging

DNA provides mechanisms which enable the event information generated by entities to be
distributed to points in the network where it can be stored or analyzed. DNA Phase V
event logging is an extension of the facilities available in Phase IV, and consists of the
following components:
• Sources
• Sinks
• Event dispatchers (EVD)

Event sources are the DNA entities that detect events and initiate the production of event
reports. Event sinks are the consumers of event information. They are entities that process
this information, or simply store or display it. Phase V event logging allows an arbitrary
number of sinks, possibly with each providing different features.
An event dispatcher, or EVD} serves as an intermediary between sources and sinks.
Each node in the network, if it reports events, has an EVD. Entities in the node post
event records to the local EVD which is responsible for buffering the information and
distributing it to event sinks. Network management can control the operation of an event
dispatcher. Event streams may be created which define the sinks for event information,
and event filters which define the event subset reported to each sink. Further filtering of
events may be carried out at the sink. This permits certain event sinks to be restricted to
the processing of particular events. Event generation may be turned off at the source. The
Management Event Notification protocol is used to carry event reports between an EVD
and an event sink.

10.2.3

Maintenance Operations Protocol (MOP)

Maintenance operations are special, primitive functions that must be available in a system
without the services of the higher layers of DNA, for example if the system is initializing
and not fully operational. DNA defines a simple management protocol for use in these
circumstances called the Maintenance Operations Protocol} or !'viOP. MOP is a client of
the Data Link layer, requiring only minimal data link protocol support; the MOP modules
handle all message acknowledgment, time-out and retransmission functions. The following
maintenance functions are defined:

Down-line loading. The down-line load function allows a memory image to be sent to a
directly connected, or adjacent system. If the image is that of a program, the function
allows program execution to be started at a specified memory address following the

CHAPTER 10. SETHiORK 1\1ANACE.\1EST

load. On a CSMA/9D data link, a system can multicast its load request and select
the first system that responds as the server.

Up-line dumping. The up-line dump function allows the contents of memory to be sent
to an adjacent system. On a CS.\tlA/CD data link~ a system can multicast its dump
request and select the first system th at responds as the sener.
Link loopback testing. MOP tests communication links by looping a test message at
various points in the physical connection. By moving the loopback point and isolating
components, a manager can diagnose link problems.
System console control. MOP can be used to control remote~ possibly unattended systems through emulation of a console terminal. This allows the remote system to be
restarted using a "Boot" command.

10.2.4

Network Control Language (NCL)

DNA defines a command line interface to directors, called NeL. NCL is new to DNA
Phase V, replacing the Network Control Program (NCP) of Phase IV. NCL gives access
to the directives defined for all of the entities in DNA.
NCL specifies the general syntax rules for network management commands, and for
display of subsequent responses. An implementation of NCL takes input from a terminal
. and causes directives to be issued to entities using MICE protocol messages. The NCL
command syntax consists of a verb, entity name and a sequence of argument identifiers or
identifier-value pairs. The following examples illustrate the syntax:
Set Node Engineering.Reading Routing Circuit DMC-l Ll Cost 10
Show Node Engineering.Reading DDCMP Link DMC-l Counters
NCL permits the use of wildcarding at various points in the command. This can be used
to effect a management operation at rnany instances of an entity class on a single node.
For example, if DMC-l was replaced by * in the Show command above, all DDCMP Link
entities on node Engineering. Reading would return their Counter values.
NCL provides commands that invoke calls to the DNA Naming Service, for example to
register a node in the namespace. Additionally, NCL permits management of a Phase IV
system from a Phase V system. Phase IV style NCP commands are introduced by the
keyword "NCP" .

Chapter 11

DNA Applications
DNA supports a wide range of applications, from general purpose network-wide applications such as file transfer and virtual terminals, to highly specific distributed applications
written to solve particular customer problems. Digital currently offers a large number of
widely-used application protocols with DNA. This chapter briefly describes some of those
application protocols. Note that all Phase IV applications will continue to run on Phase V
systems.

11.1

Heterogeneous File Access and Transfer

DNA provides the Data A ccess Protocol (DA P) for accessing and transferring files in a
heterogeneous DECnet network. DAP provides the following functions and features:
• Supports heterogeneous file systems.
• Retrieves a file from an input device (disk file, terminal, etc.).
• Stores a file on an output device (magtape, printer, disk file, etc.).
• Transfers files between systems.
• Supports creation, deletion, and renaming of remote files.
• Lists directories of remote files.
• Recovers from transient errors and reports fatal errors to the user.
• Allows multiple data streams to the same remote file.
• Submits and executes remote command files.
• Permits sequential, random, and indexed (ISAM) access of records.
• Supports wildcard file specification for sequential file retrieval, file deletion, file renaming, and command file execution.
87

CHAPTER 11. DNli APPLICATIONS

• Permits an optional file checksum to ensure file integrity.
DAP is designed to minimize protocol overhead. For example~ the file transfer mode
eliminates the need for DAP control messages once a data file flow begins. Also, small file
records can be blocked together and sent in one large message.
Two cooperating processes exchange DAP messages: the user process and the server
process that acts on the user's behalf at the remote system. C ser I/O commands accessing
a remote file are mapped into equivalent DAP messages and transmitted via a transport
connection to the server at the remote system. The server interprets the DAP commands
and actually performs the file I/O for the user. The server returns status and data to the
user.

11.2

Network Virtual Terminals

The Network Virtual Terminal Service consists of the following:
• A functional model of a terminal (virtual terminal) for access through the network.
This model is known as the network command terminal.
• A set of protocols for communicating between a host system, with an application
program running, and a server system, with a terminal directly attached.
The Network Virtual Terminal Service provides these functions and features:
• Distributes terminal handling functions between two systems.
• Supports heterogenous host systems. Different operating systems can cooperate via
common protocols, and a host operating system can manage a terminal in its own
way, regardless of which operating system runs in the server.
• Allows a server to connect to a specified host or a host to a specified remote terminal.
• Provides terminal input/output and characteristics management functions at the
operating system services level. These include:
Accept input even if the program has not issued a read request (typeahead).
Take action on certain characters immediately as keys are struck (out-oJ-band
character processing).
Recognize ANSI standard escape sequences on input and output.
Read and set terminal device characteristics.
• Offers standard terminal services, featuring good performance and device independence. Optionally, offers methods of controlling the terminal's behavior in considerable detail.

11.3. ELECTRONIC MAIL

The virtual terminal service is realized by a pair of protocols, CTERM and FOUND.
FOUND provides connection management and a transparent data transport capability
over which a number of terminal usage models may co-exist (for example: command mode,
editor mode, forms mode). CTERM implements the model of a command terminal which
is the common mode of access to command language processors such as DCL (Digital
Command Language).

11.3

Electronic Mail

11.3.1

Mail-II

The Mail-II protocol is a DNA application protocol which provides a personal electronic
mail capability. It allows people to exchange text messages among VMS™, ULTRIX™,
RSX-IIM-PLUS™, and RSTS/E™ systems. The Mail-II protocol is implemented on
VAX/VMS™ by the Mail utility, which also handles the user interface to mail on the local
host. Mail-II is implemented on RSTS/E and RSX-IITM via the DECMail-ll™ product.
Gateway products provide interfaces from the Mail-II protocol to other protocols including:
• VAX PSI Mail, which permits mail to be sent between distinct DNA networks using
the VAX PSI X.25 communications product.
• Message Router™ (see Section 11.3.2), which permits users of the VAX/VMS Mail
utility to gain access to the enhanced capabilities of Message Router including the
ability to send mail using the X.400 suite of protocols.
• Ultrix Mail, which permits mail to be sent not only to Ultrix systems in the same
DNA network, but also to other systems utilizing the TCP /IP SMTP Mail protocols
and UUCP mail.

11.3.2

Message Router

The Message Router, together with its family of layered products, provides a highly reliable
store and forward message handling system for the delivery of interpersonal mail based
upon the X.400 series of international standards for message handling systems.
Included in the Message Router product family are mail user agents, which provide
message transmission and delivery services to users, message transfer agents, which provide
store and forward message transport in a DECnet network, and gateways, which provide
access to other mail systems and other networks.
Together these components provide common interpersonal mail services to users of
multiple vendors' computer systems interconnected by multiple network products.

11.4

SNA Interconnect Applications

DNA provides a rich set of application protocols for communicating with systems conforming to IBM's Systems Network Architecture (SNA). The following subsections briefly

CHAPTER 11. DNA APPLICATIONS

describe some of these SN A-orien LeJ applications.

11.4.1

SNA Gateway Access

SJ\A Gateway Access provides the following functions and features:
• Supports communication between user-written and Digital-written programs in a
host DECnet system, and modules in an IBM host system~ over an S\,A network.
• Supports programs in DECnet systems which act as SNA Secondary Logical enits
(SLUs) of any of the defined SNA Logical Unit Types (LUs).
• Allows user programs full access to the facilities provided by the SN A Transmission
Control and Data Flow Control layers.
• Communicates with the SNA network as a Physical Unit Type 2 (PU2).
• Permits user-written programs to participate in SNA sessions from any DECnet system containing an SNA Gateway Access module. The user program does not have
to reside in the DECnet system that is acting as the PU2 directly connected to the
SNA network.
• Permits the user-written program to send messages to and receive messages from the
SNA System Services Control Point - that is, the user program has access to the
SSCP-SL U session.
• Manages the SNA session pacing automatically for the user program.
• Allows an IBM application to initiate a session with a DECnet partner, by automatically activating the DECnet object when a session request (BIND message) IS
received from the SNA network.
• Provides an access control mechanism to restrict SNA session address use to authorized users.
• Recovers from transient errors and reports fatal errors to the user.
To use the services of SNA Gateway Access, the user program makes calls to the SNA
Gateway Access module. SNA Gateway Access communicates with a process called the
SNA Gateway Server which runs in a DECnet system that is also a PU2 node on the
SNA network (the "gateway system"). The communication occurs over DECnet Session
Control; SNA Gateway Access Protocol messages are exchanged over the connection to
make the facilities of SN A sessions available remotely. Figure 11.1 illustrates the operation
of SNA Gateway Access.
SNA Gateway Access provides the basis for a range of Digital products, including
3270 terminal emulation, printer emulation, remote job entry, DISOSS data exchange, file
transfer and data access, and various user programming interfaces.

11.4. SNA INTERCONNECT APPLICATIONS

LOCAL SYSTEM
User Layer
The accessing program

(SLU):
SNA Remote Job Entry
IBM HOST SYSTEM

3270 Emulation

PLU

DTF/VMS
User-written program

GATEWAY SYSTEM

DNA Application

SNA Gateway Access
Module

SNA Gateway Server
Module

Session Control
Transport
Network
Data Link
Physical

The cooperating program:
(e.g. DTF /MVS, CICS,
IMS, JES2, JES3, etc.)
NAU Services
Data Flow Con trol
Transmission Control
Path Control
Data Link
Physical Link

Path Control

(PU2)
Data Link
Physical Link

.......

......

DECnet Network

.......

SNA Network

Legend:

SL U PLU P U2 P U4 -

Secondary Logical Unit
Primary Logical Unit
Physical Unit type 2
Physical Unit type 4

---z---. A physical communication link
................

A path through a network

Figure 11.1: SNA Gateway Access Operation

CHAPTER 11.

11.4.2

DSA API)LICATIOSS

3270 Terluinal En1ulator

The 3270 Terminal Emulator (TE) mimics an IB\rl 327 x block-mode terminal connected to
a 3274 cluster controller. and thereby allows a YAXY!vlS user at a VT100TM: VT200nv~:
or VT300-class terminal 1 to interactively log on to IB\'l host applications. The TE handles
the conversion to and from the data formats used by 3270-class terminals, manages the
efficient updating of the VT terminal screen. and communicates 'with the IB\l host as an
S!\A Logical "Cnit Type 2 through the S:\A Gateway.

11.4.3

Remote Job Entry

The Remote Job Entry subsystem provides batch access to IBM's MVS operating system.
An RJE server, resident in the gateway itself, supports emulation of up to four workstations.
VAX/VMS users can submit J CL files to a queue on a host VAX associated with one
of the workstations; the RJE server will eventually transmit the JCL "deck" to the IBM
mainframe for processing. Later, print and punch output from the job may be routed back
to any system in the DECnet network (the server uses ordinary DECnet remote file access
to move its input and output data around).
A VAX/VMS operator utility provides access to the console of the emulated RJE
wor kstation.

11.4.4

Data Transfer Facility

The Data Transfer Facility (DTF) software provides for bidirectional file transfer between
VAX/VMS and IBM's MVS system. File transfers can be requested from the VAX and
from the IBM machine. In addition, user-written programs in the VAX/VMS system can
access IBM files at the record level, using standard RMS calls.
The DTF software includes VMS-resident components (DTF /VMS) and MVS-resident
components (DTF /MVS); both parts are written and supplied by Digital. In the VMS
system, DTF /VMS is implemented as a data transfer server which can be used by any client
in the DECnet network which supports DAP (see Section 11.1). DTF /VMS communicates
with DTF /MVS through the SNA Gateway using the DECnet DAP protocol running on
top of a private Logical Unit Type 0 protocol. Thus, unlike earlier S1'\A Gateway layered
products which emulated existing IBM devices, DTF implements a DECnet protocol on
the IBM system itself.
DTF /VMS also provides the ability to translate the data records during transfer according to field definitions stored in the Common Data Dictionary (VAX CDD™). Thus,
records containing mixed data can be accommodated, as well as simple EBCDIC to ASCII
translations.
lOr a compatible terminal emulator, such as a workstation.

11.5. VMS SERVICES FOR MS-DOS

11.5

VMS Services for MS-DOS

VMS Services for MS-DOS is a VAX based remote file server for personal computer systems
running Microsoft's MS-DOS operating system. The server allows personal computers,
running Digital's Personal Computer Systems Architecture (PCSA) software, to create,
read and modify MS-DOS files on a VMS system. The user connects to VMS services
(established by the VAX system administrator), via the PCSA client system software, and
is presented with the equivalent to a MS-DOS file oriented disk device. File sharing and
byte range locking are implemented (according to the Microsoft specification) by the file
server.
The server implements the Microsoft Networks (MS-NET) Server Message Block (SMB)
protocol, version 1.0. The server provides an MS-NET session level interface; all lower layers (Transport, Network, etc.) are provided by DECnet. The software, although intended
to operate in aLAN, will also operate in a WAN environment. The file server may be
installed on stand-alone VAX systems, or on VAXcluster™ systems.

11.6

Time Service

The notion of time is taken for granted in most of today's centralized computer systems.
However, the mechanisms used for providing time in these systems are inadequate when
applied to distributed systems. In the distributed environment, new mechanisms are required. A global notion of time can be extremely useful to co-ordinate the operation of a
distributed system or distributed applications.
The Digital Time Service is an architecture for providing and maintaining correct time
in a distributed system. The architecture conforms to the client-server model of distributed
systems whereby a collection of time servers supply the time to clients through intermediaries, called clerks, residing on the same machines as the clients. Since the value of
the time at any instant is never known exactly, the time can at best be expressed in the
form of an interval which is guaranteed to contain the instantaneous value of the time.
By obtaining the time from a number of servers, systems are able to compute a correct
time interval even in the presence of some faulty servers. This ensures that the time obtained by clients is correct with arbitrarily high probability. The Time Service architecture
specifies the operation of clerks and servers including the methods for obtaining the time
from a server, computing the best time interval from a set of intervals, maintaining the
local clock correctly, and detecting faulty servers. The algorithms are fully distributed and
asynchronous.

11.7

Computer Conferencing

VAX Notes Computer Conferencing allows people throughout a DNA network to participate in round-table discussions, bridging distances between the participants and accomodating time differences in working hours in a global network to enhance information sharing
throughout an organization.

CHAPTER 11. D1'.IA APPLICATIOIVS

Any number of VAX ::~otc; Conferences can be established in a nptw()rk. Conference
participants use the VAX :\otes client program installed on their home systems. These
client programs communicate with VAX Notes server programs using a DI\A application
protocol which allows access to conferences located throughout the network for reading
and writing conference entries, known as notes.
Conference moderators create conferences for the discussion of a major topic or group
of related topics. Each conference can be considered to be logically equivalent to a large
bulletin board organized into individual topics, which. continuing the bulletin board analogy, are posted along the top of the bulletin board by the participants. Posting a note at
the top of the bulletin board is known as writing a new topic. Participants who wish to add
to the discussion of existing topics "reply" to that topic; in the bulletin board analogy this
is equivalent to posting a note directly below the existing topic and any existin g replies.
The moderator of a conference has special capabilities for controlling the discussion,
including the ability to restrict conferences to a membership list, to delete or "hide" notes
deemed inappropriate to the discussion or which require further clarification, to change the
titles of topics or replies to topics to improve the organization of a conference, to create
keywords which may be associated with notes in the conference, and to designate additional
moderators.

11.8

VAX System Performance Monitor

The VAX System Performance Monitor (VAX SPM™) uses a DNA application protocol
to allow a single system in a VAXcluster ™ system to communicate with all other members
of the cluster. VAX SPM uses this for starting and stopping data collections on the entire
cluster from a single system and for doing live video displays on the cluster.
The VAX SPM DNA application works asynchronously. When performing a cluster
command, a connection is established to each system in the cluster. The protocol establishes the links, and sends a request for data to all of the other systems. The other systems
send the data and when all the data has been sent, an end of data message is sent. The
receiving end keeps track of which systems have completed their data transmission and
when all systellls have completed, the data is processed.

11.9

Videotex

Videotex is a computer based electronic retrieval system which uses a hierarchical arrangement of information, is video-terminal based, and provides a very simple user interface.
VAX VTX™ is Digital's product which provides videotex services on VAX/VMS systems.
VAX VTX uses a DNA application protocol for communication between participating systems in a DNA network. VAX VTX organizes information into pages, allowing pages to
be retrieved from in/abases, which can be resident on any VAX/VMS system.
VTX users, known as subscribers, can display, print, and save VTX pages using simple
keypad techniques. Subscribers use a client program, known as a pad (packet assembler / disassembler), which runs on their system and communicates with a server process

11.10. DISTRIBUTED QUEUEING

on the system where the info base resides. The server provides access to the infobase files,
making the infobase available for perusal either by general users, or by specified groups as
determined by parameters specified on the server system.
The page information itself can be stored in a variety of display protocols, such as
ReGIS™, ASCII, NAPLPS, and PRESTEL, giving VTX a considerable amount of flexibility. VAX VALU™, the VTX sister product, provides limited data input and collection
via the use of form pages.

11.10

Distributed Queueing

The Vax Distributed Queuing Service (DQS) extends the VMS queue system and increases
the availablity of printing services in a distributed environment. DQS allows files to be
printed on remote VMS systems which have printing devices attached. DQS also supports
the examination and manipulation of print jobs.
DQS can be configured to operate as both a client and a server or as a client only. Any
system where a print request originates is called a client. Any system with the attached
printer that performs the printing is called a server. Client or server systems may be added
in the network at any time without disrupting current operations.

11.11

Remote System Management

The Remote System Manager is used within a DECnet network to aid in the management
of VMS and ULTRIX systems. Services provided by RSM include:
• File system backup and restore
• Software distribution and installation
RSM operates using a client/server model. The server is a VMS system which contains
the backup schedule and software libraries. A manager, located on the server, schedules
backups and software distribution operations for one or more client systems.
RSM uses a DNA application protocol to send commands to an agent located on each
VMS or ULTRIX system being managed. Upon receiving instructions from the server, the
client's agent may perform a backup using the local backup utility, or install an application.
RSM also uses the DNA MOP protocol (see Section 10.2.3), combined with special RSM
applications, to load either the VMS or ULTRIX operating system onto a client system
over a local area network.
RSM stores information about RSM clients and servers using the Naming Service, which
provides RSM with a network-wide namespace (see Chapter 8).

CHAPTER 11. DNA APPLICATIONS

Chapter 12

OSI Upper Layers and
Applications
As the DNA Transport layer and below contain protocols which are standardized as part
of OSI, OSI applications can be built which interwork with other open systems.
In DNA Phase V, OSI upper layer components coexist with existing DNA higher layer
components as a distinct set of layers and share access to the Transport layer with the
DNA Session Control layer (see Figure 1.3). OSI upper layers provide communication for
OSI applications. OSI upper layers are not integrated into DNA Phase V as the standards
describing them are not yet complete.
The components of the upper layers of OSI fall into two categories: those which are general building blocks from which complex OSI applications can be built, and those which are
less general or provide the communications facilities to fulfill specific information processing requirements. Section 12.1 discusses the former components and Section 12.2 discusses
the latter.

12.1

General OS1 Upper Layer Components

Together, the upper layers of OSI provide the communications services for OSI applications
partaking in distributed information processing activities. The services of a given layer are
made visible by pass through services to the next higher layer, and so on up through the
layers. Each layer also contributes services pertaining to its layer functions. The totality
of services passed through from the OSI Session and Presentation layers, together with
specific services provided in the Application layer are provided to OSI applications. The
OSI upper layers contribute the following general purpose functions:
Session layer. The facilities of the OSI Session layer build on the reliable end to end communication facilities provided by the DNA Transport layer. They provide a means by
which communicating OSI applications can control access to certain services, allow
synchronization and resynchronization to specific points in the communication, and
allow the division of the communication into a number of logical pieces of work.
97

CHAPTER 12. 051 UPPER LA YERS AIVD APPLICATIO.rVS·

D\"A applications~ which require facililie~ ~iIIlilal to those provided by the OS1 Scs
sion layer, meet their requirements in an application specific way.
Presentation layer. The facilities of the OS1 Presentation layer build on and pass through
the services provided by the OS1 Session layer and provide, in addition, a means by
which OS1 applications can agree on representations for data while in transit between communicating partners. Subsequently, the Presentation layer causes the data
transferred to be in the agreed representations. This provides OS1 applications with
a means to transfer information without loss of the application semantics.
These facilities are required in order to support the interworking of applications in
heterogeneous networks where the local representation of data may differ between
each open system. DNA applications do not require these facilities as they only
operate in homogeneous networks.
Application layer. The facilities of the Application layer are subdivided into a number of
application service elements (AS E) , each of which provided a specific set of functions
in a reusable way. Some ASEs form the components of specific applications and are
described in Section 12.2. General purpose ASEs are described below:
Association Control Service Element. Association control services are provided
by the Association Control Service Element (ACSE). These allow OSI users to
establish an application association between them for the purpose of exchanging
information. When information exchange is complete, ACSE provides services
to destroy these associations.

12.2

OSI Applications

Several OSI applications are currently under development as part of the OSI set of standards. These standards define the information processing tasks required and the application
service elements necessary to fulfill the communications related requirements.
The following specific applications are of interest as they provide similar functions to
those provided by DNA applications:
File Transfer, Access and Management Services. FTAM is a four part International
Standard which provides the following services:
• transfer of an entire file;
• read, write or modify of existing components (that is, records) of a file;
• file management - creation of files, deletion of files and reading and modification
of file attributes.
When information is transferred between open systems, FTAM preserves the semantics and structure of the file data being exchanged.

12.2. OSI APPLICATIONS

Virtual Terminal Services. VT consists of a service definition and a protocol specification which provides services that are similar to those provided by the DNA Network
Virtual Terminal (NVT) service but in a heterogeneous environment. Specifically, it
supports the interactive transfer and manipulation of structured data in a manner
which models character cell terminals.

100

CHAPTER 12. 051 UPPER LAYERS AND APPLICATIONS

Glossary
access control

The management of rights to use resources.

address prefix

Any leading portion of an NSAP address.

address resolution The Session Control function which maps from a Naming Service
object name to the identifiers of protocols and corresponding addresses which are mutually supported by the local system and the remote system(s) on which the named object
resides.
address selection The Session Control function which provides transparent selection of
protocols and addresses for Transport Connection establishment based upon the destination object name.
addressing authority The authority responsible for the unique assignment of Network
layer addresses within an addressing domain.
addressing domain A level in the hierarchy of Network layer addresses. Every NSAP
address is part of an addressing domain that is administered directly by one and only
one addressing authority. If that addressing domain is part of a hierarchically higher
addressing domain (which must wholly contain it), the authority for the lower domain
is authorized by the authority for the higher domain to assign NSAP addresses from the
lower domain.
administrative domain A collection of End Systems, Intermediate Systems and Subnetworks operated by a single organization or administrative authority. It may be subdivided into a number of routing domains.

AFI

See Authority and Format Indicator.

agent

That part of an entity which provides the interface to network management.

area

The group of systems which constitute a single Levell routing subdomain.

area address The concatenation of the lOP and LOC-AREA fields of an NSAP address.
A system may have more than one area address, but the systems making up an area
must have at least one area address in common with each of their neighbors.
101

GLOSSARY

102

asynchronous link A communications link using asynchronous transmission, in which
each transmitted character is framed with start and stop bits. The time interval between
characters may be of unequal length.

Authority and Format Indicator The part of an NSAP address which indicates the
addressing authority responsible for the assignment of the IDP and its format. It also
indicates the format (binary or decimal) of the DSP.
balanced mode

An HDLC operational mode used in DNA over full duplex links.

bridge A Data Link relay for interconnecting LANs, used to increase the maximum
number of stations, maximum distance, and total available bandwidth.
broadcast subnetwork A multiaccess subnetwork that supports the capability of addressing a group of attached systems with a single message.
call reference

A unique value used locally to identify a Modem Connect or X.21 call.

call sharing A form of switched line sharing in which many clients have access to the
same call on that line.
carrier sense On CSMA/CD LANs, each station waits before transmitting for the carrier indication from the Physical layer to go off, indicating that no other station is in
the process of transmitting.
CCITT The International Telegraph and Telephone Consultative Committee, the technical committee of the International Telecommunications Union (ITU), responsible for
the development of recommendations regarding telecommunications, including data communications.
centralized management A form of network management where management is performed from a single point in the network.
checksum A value transmitted with a message which is computed from the content of
the message itself. This value is recomputed by the receiver of the message to detect
whether the message was corrupted during transmission.
child directory entry An entry in the Naming Service which points to a child directory
of some directory in the namespace.
circuit switching A telecommunications technique involving the dynamic establishment
of a physical connection prior to information exchange, and release of the connection
following the exchange.
clearinghouse

A collection of directory replicas stored together in one location.

client

The user of the service provided by a module or layer in the architecture.

CLNS

See Connectionless-mode Network Service.

103

GLOSSARY
CMIP

See Common Management Information Protocol.

collision An attempt by two stations on a CSMA/CD LAN to transmit at the same
time. The stations will detect the collision and retry the transmission.
Common Management Information Protocol
compasses the MICE and MEN protocols.
communications link
concatenation
transmission.

A management protocol which en-

The physical medium connecting two systems.

The process of assembling multiple nPDUs into a single (n - l)SDU for

congestion Congestion occurs when a network, or part of a network, is overloaded and
has insufficient communication resources for the volume of traffic.
congestion avoidance
congestion.

A mechanism used to adjust the load on the network to prevent

connection control The Session Control function concerned with the system-dependent
functions related to creating, maintaining, and destroying Transport Connections.
Connection-mode Network Service A network service which operates according to
a connection oriented model. Before data can be exchanged, a connection must first be
established.
Connectionless-mode Network Service A network service which operates according
to a datagram model. Each message is routed and delivered to its destination independently of any other. The Network Layer of DNA provides this type of service.
CONS

See Connection-mode Network Service.

control station
link control.

In DDCMP multipoint configurations, the station responsible for data

cost A metric used by the routing algorithm. Each link is assigned a cost, and the routing
algorithm selects paths with minimum cost.
credit A flow control mechanism whereby the receiver of data tells the transmitter how
many messages it is prepared to receive at a given point.
credit window When operating a credit based flow control mechanism, the credit window identifies the range of message numbers which the receiver is prepared to receive at
a given point.
CSMA/ CD Carrier Sense Multiple Access with Collision Detection, the channel access
method used by Ethernet and ISO 8802-3 LANs.

GLOSSARY

104

cyclic redundancy check (CRC) An error detectlOn mechanism which involves polynomial division of the data to be transmitted. The coefficients of the remainder are
transmitted along with the data: for checking by the receiver.
DA

See Dynalnic Assignment.

DAP

See data access protocol.

data access protocol A D:\A proprietary application protocol for accessing and transferring files across a DNA network.
data circuit equipment The equipment that provides the functions required to establish: maintain and terminate a connection between DTEs using a physical circuit or a
virtual circuit.
data link connection
connected systems.

The Data Link layer communications path between two directly

Data Link layer The layer of DN A which provides a communications path between two
directly-connected systems in a network.
data link protocol The coding rules and communication procedures which allow the
data link service to be provided.
data terminal equipment The equipment, typically a computer system or terminal,
comprising a data source and sink connected to common carrier communication facilities.
datagram A frame that is processed by the network independent of any other frames.
Datagrams may fail to be delivered without notice.
DCE

See data circuit equipment.

DCM

See Dynamic Connection Management.

DED

See Dynamically Established Data Link.

Digital Data Communications Message Protocol (DDCMP)
protocol used in DNA.

A Data Link layer

directive

A management request sent by a director to an entity.

director

The management software used by a network manager.