Digital PDFs

XX-B6A3B-B4

December 2000

70 pages

Original

0.1MB

Document:	ds5500 tpcb informix dec3
Order Number:	XX-B6A3B-B4
Revision:
Pages:	70
Original Filename:	https://web-docs.gsi.de/~kraemer/COLLECTION/ULTRIX/ds5500_tpcb_informix_dec3.pdf

OCR Text

TPC Benchmark B
Full Disclosure Report
for the
DECsystem 5500
Using
ULTRIX 4.2 and INFORMIX-OnLine 4.10

System
Name

Company
Name

Digital Equipment DECsystem 5500
Corporation

Total System Cost

-Hardware
-Software
-5 years Maintenance

$160,113

Database
Software

Operating System
Software

INFORMIXOnLine 4.10

ULTRIX 4.2

TPC-B Throughput

Price Performance

Sustained maximum throughput of system running TPC
Benchmark B expressed in
transactions per second.

Total system cost/
TPC-B throughput
($160,113/40.6 tps-B)

40.6 tps-B

$3,944 per tpsB

TM
Submitted for Review: 12/3/91

First Printing December, 1991
Digital Equipment Corporation believes that the information in this document is accurate as of its
publication date. The information in this document is subject to change without notice. Digital
Equipment Corporation assumes no responsibility for any errors that may appear in this document.
The pricing information in this document accurately reflects prices in effect on the indicated dates.
However, Digital Equipment Corporation provides no warranty on the pricing information in this
document.
The performance information in this document is for guidance only. System performance is highly
dependent on many factors including system hardware, system and user software, and user application characteristics. Customer applications must be carefully evaluated before estimating performance. Digital Equipment Corporation does not warrant or represent that a user can or will
achieve similar performance expressed in transactions per second (TPS) or normalized
price/performance ($/TPS). No warranty on system performance or price/performance is expressed or implied in this document.
Copyright ©1991 Digital Equipment Corporation
All Rights Reserved.
Printed in U.S.A.

DEC, DEC C, DECsystem 5500, ULTRIX and the DIGITAL Logo are trademarks of Digital Equipment Corporation.

Informix is a registered trademark of Informix Software, Inc.
Prestoserve is a registered trademark of Legato Systems, Inc.
TPC Benchmark and TPC-B are a trademarks of the Transaction Processing Performance Council.

Abstract
This report documents the compliance of testing performed on a
DECsystem 5500 server running INFORMIX-OnLine 4.10 in conformance to the Transaction Processing Performance Council Benchmark B
Standard Specification.
Two standard metrics, transactions per second (TPS) and price per TPS
($/TPS), are reported. Throughout this report, TPS refers to the tpsB performance metric, in accordance with the TPC Benchmark B Standard.
The independent auditor’s report by KPMG Peat Marwick is included.

iii

Table of Contents

Preface

................................................................................................................ viii

TPC Benchmark B Full Disclosure

1 - General Items ................................................................................................................ 1
1.1

Sponsor ................................................................................................................................. 1

1.2

Application Code and Definition Statements ....................................................................... 1

1.3

Parameter Settings ................................................................................................................ 2

1.4

Configuration Diagrams ....................................................................................................... 2

2 - Clause 2 Related Items ............................................................................................... 3
2.1

Atomicity Tests .................................................................................................................... 4

2.2

Consistency Tests ................................................................................................................. 5

2.3

Isolation Tests ...................................................................................................................... 5

2.4

Durability Tests .................................................................................................................... 7

3 - Clause 3 Related Items ............................................................................................... 8
3.1

ABTH Data Storage Distribution ......................................................................................... 8
3.1.1
History Storage and Recovery ........................................................................... 10

3.2

Database Contents and Method of Population ................................................................... 13

3.3

Type of Database ................................................................................................................ 13

4 - Clause 4 Related Items ............................................................................................. 13
5 - Clause 5 Related Items ............................................................................................. 13
5.1

Method of Verification of Random Number Generator ..................................................... 13

5.2

Horizontal Partitioning Disclosure ..................................................................................... 13

6 - Clause 6 Related Items ............................................................................................. 14
7 - Clause 7 Related Items ............................................................................................. 16
7.1

Determining Steady State ................................................................................................... 16

7.2

Work Performed During Steady State ................................................................................ 17

7.3

Determining Reproducibility ............................................................................................. 19

7.4

Duration of Measurement Period ....................................................................................... 20

8 - Clause 8 Related Items ............................................................................................. 20
8.1

Description of the Driver ................................................................................................... 20

9 - Clause 9 Related Items ............................................................................................. 21
9.1

Hardware and Software Components ................................................................................. 21
9.1.1
Priced System Configuration Tables ................................................................. 21
9.1.2
Package Pricing ................................................................................................. 22

9.2

Total Price of System Configuration .................................................................................. 23
9.2.1
Hardware Pricing ............................................................................................... 23
9.2.2
Software Pricing ................................................................................................ 23
9.2.3
Price Discounts .................................................................................................. 23
9.2.4
System Pricing Summary .................................................................................. 25

9.3

Performance and Price/Performance .................................................................................. 26

10 - Clause 10 Related Items ........................................................................................ 26
11 - Clause 11 Related Items ........................................................................................ 26
11.1

Independent Auditor’s Report ............................................................................................ 26

Appendix A Application Code ............................................................................... A-1
A.1

tpc.ec source code ............................................................................................................ A-1

A.2

createdb.ec source code .................................................................................................... A-6

A.3

createhist.ec source code .................................................................................................. A-8

A.4

createruns.ec source code ............................................................................................... A-10

A.5

createidx.ec source code ................................................................................................. A-13

A.6

config.scr source code .................................................................................................... A-14

A.7

bench.h code ................................................................................................................... A-15

Appendix B Database Definitions ......................................................................... B-1
Appendix C Code to Populate Database ........................................................... C-1
C.1

Database Population Program .......................................................................................... C-1

Appendix D Database Contents Samples ........................................................ D-1

D.1

Branch Table .................................................................................................................... D-1

D.2

Teller Table ...................................................................................................................... D-1

D.3

History Table .................................................................................................................... D-2

D.4

Account Table .................................................................................................................. D-2

Appendix E Device Configurations ..................................................................... E-1
Appendix F System Parameter Settings ........................................................... F-1
F.1

System Parameters ............................................................................................................ F-1

F.2

IPC Semaphore Facility .................................................................................................... F-3

Appendix G Independent Auditor’s Report ..................................................... G-1

vii

Preface
This report documents the compliance of the Digital TPC Benchmark B testing on a
DECsystem 5500 with the TPC Benchmark B Standard Specification1. The TPC
Benchmark B Standard represents an effort by Digital Equipment Corporation,
Informix Software Inc., and other members of the Transaction Processing Performance Council (TPC) to create an industry-wide benchmark for evaluating the performance and price/performance of transaction processing systems.
These tests were run using the INFORMIX-OnLine relational database running under
the Digital ULTRIX operating system.
Document Structure
The TPC Benchmark B Full Disclosure Report is organized as follows:
•

The main body of the document lists each item in Clause 10 of the TPC Benchmark B Standard and explains how each specification is satisfied.

•

Appendix A contains the source code of the application program used to implement the TPC Benchmark B transaction and related programs and scripts.

•

Appendix B contains the INFORMIX-OnLine database definitions.

•

Appendix C contains the source code used to populate the database.

•

Appendix D contains samples of contents of the database files used in the tests.

•

Appendix E contains a description of the physical disk partitions.

•

Appendix F contains the operating system parameters and options.

•

Appendix G contains the Independent Auditor’s Report by KPMG Peat Marwick.

Additional Copies
To request additional copies of this report, write to the following address:
Digital Equipment Corporation
Administrator, TPC Benchmark Reports
Transaction Processing Systems Group
151 Taylor Street (TAY1)
Littleton, MA 01460-1407
U.S.A.
FAX number: (508) 952-4197

1 TPC Benchmark B Standard Specification, Transaction Processing Performance Council, August 23, 1990,
and addenda as of September 20, 1991.

viii

TPC Benchmark B Full Disclosure
The TPC Benchmark B Standard Specification requires test sponsors to publish,
and make available to the public, a full disclosure report in order for the results to be
considered compliant with the standard. The required contents of the full disclosure
report are specified in Clause 10.
This report is intended to satisfy the TPC Benchmark B standard’s requirement for
full disclosure. In the TPC Benchmark B Standard Specification, the main headings in Clause 10 are keyed to the other standard clauses. The headings in this report use the same sequence, so that they correspond to the titles or subjects referred
to in Clause 10.
Each section in this report begins with the text of the corresponding item from
Clause 10 of the TPC Benchmark B Standard Specification, printed in italic type.
The plain type text that follows explains how the tests comply with the TPC Benchmark B requirement. In sections where Clause 10 requires extensive listings, the section refers to the appropriate appendix at the end of this report.

1 - General Items
1.1 Sponsor
A statement identifying the sponsor of the benchmark and any other companies who
have participated.
This benchmark test was sponsored by both Digital Equipment Corporation and
Informix Software, Inc. The results were attested to by KPMG Peat Marwick.

1.2 Application Code and Definition Statements
Program listing of application code and definition language statements for
files/tables.
•

Appendix A contains the C source code of the application program used to implement the TPC Benchmark B transaction and related programs and scripts.

•

Appendix B contains the INFORMIX-OnLine database definitions.

•

Appendix C contains the source code used to populate the database.

•

Appendix D contains samples of contents of the database files used in the test.

•

Appendix E contains a description of the physical disk partitions.

•

Appendix F contains the operating system parameters and options.

•

Appendix G contains the Independent Auditor’s Report by KPMG Peat Marwick.

TPC Benchmark B Full Disclosure 1

General Items

1.3 Parameter Settings
Settings for all customer-tunable parameters and options that have been changed
from the defaults found in actual products; including but not limited to:
•

Database options

•

Recovery/commit options

•

Consistency/locking options

•

System parameters, application parameters, and configuration parameters

Test sponsors may optionally provide a full list of all parameters and options.
A listing of all parameters and options is provided.
Appendixes A, B, E, and F contain the application, database configuration, partition,
and operating system parameters used in the TPC Benchmark B tests.

1.4 Configuration Diagrams
Configuration diagrams of both benchmark configuration and the priced system, and
a description of the differences.
The configurations used for the benchmark and the priced system were the same.
The configuration consisted of a DECsystem 5500 with 32 Megabytes (MB) of main
memory. Two (2) embedded controllers were utilized; a DSSI controller supporting
seven (7) 381MB RF31 disk drives and a SCSI controller supporting three (3) 665 MB
RZ56 and two (2) 1 GB RZ57 disk drives.
We enabled continuous archiving of the logical logs. The logical logs were backed up
to an archive device. Two 1 Gbyte RZ57 disk drives were used for this purpose.
These disk drives provided the necessary storage capacity so that 8 hours of log data
could be kept on-line. Informix transaction logging was at all times set to unbuffered
mode.

2 TPC Benchmark B Full Disclosure

Clause 2 Related Items

Benchmark and Priced System Configuration
The diagram that follows represents the benchmark configuration and priced system
configuration.

1 TK50
Tape Drive

DECsystem
5500
CPU

DSSI
controller
7 DSSI Disks

RF31 RF31 RF31 RF31 RF31 RF31 RF31
381Mb 381Mb 381Mb 381Mb 381Mb 381Mb 381Mb

SCSI
controller
5 SCSI Disks

RZ56 RZ56 RZ56 RZ57 RZ57
665Mb 665Mb 665Mb 1Gb

1Gb

2 - Clause 2 Related Items
ACID Properties
Results of the ACIDity test (specified in Clause 2) must describe how the requirements
were met.

Clause 2 of the TPC Benchmark B Standard lists specific tests to ensure the
atomicity, consistency, isolation, and durability (ACID) properties of the SUT (System Under Test). The following subsections show how the tests required in Clause 2
were performed. All mechanisms needed to ensure full ACID properties were enabled during both the measurement and test periods. A fully-scaled database was
used for the atomicity, consistency, and isolation tests.

TPC Benchmark B Full Disclosure 3

Clause 2 Related Items

2.1 Atomicity Tests
Atomicity of Completed Transaction

Perform the standard TPC Benchmark B transaction for a randomly selected account
and verify that the appropriate records have been changed in the Account, Branch,
Teller, and History files/tables.
The following test was performed and verified the atomicity of completed transactions:
1. Select a random Branch record.
2. Select a random Teller record.
3. Select a random Account record.
4. Count the History records.
5. Using the randomly selected records, perform the following steps:
A. Update the Branch record.
B. Update the Teller record.
C. Update the Account record.
D. Insert the History record.
E. Commit the transaction.
F. Select the Branch record.
G. Select the Teller record.
H. Select the Account record.
6. Count the History records. Verify that the History record count reflects the committed transaction.
Atomicity of Aborted Transaction
Perform the standard TPC Benchmark B transaction for a randomly selected account,
substituting an ABORT of the transaction account for the COMMIT of the transaction. Verify that the appropriate records have not been changed in the Account,
Branch, Teller, and History files/tables.
The following test was performed, and verified the atomicity of aborted transactions:
1. Select a random Branch record.
2. Select a random Teller record.
3. Select a random Account record.
4. Count the History records.
5. Using the randomly selected Branch, Teller and Account records from above, do
the following:

4 TPC Benchmark B Full Disclosure

Clause 2 Related Items

A. Update the Branch record.
B. Update the Teller record.
C. Update the Account record.
D. Insert the History record.
E. Abort the transaction and perform a rollback recovery.
F. Select the Branch record.
G. Select the Teller record.
H. Select the Account record.
6. Count the History records. Verify that the History record count has not changed.

2.2 Consistency Tests
Consistency is the property of the application that requires any execution of the transaction to take the database from one consistent state to another.
The following tests were performed and verified the consistency property of transactions:
1. Consistency of Branch and Teller records before transactions
A. Select Branch balances for each Branch record.
B. Select the sum of Teller balances for each Branch record.
C. Verify that the balance for each Branch record is equal to the balance of its
Teller records.
2. Consistency of Branch and Teller records after transactions
A. For the entire History file, count the History records and sum their delta values.
B. Perform the standard TPC Benchmark B test and record the number of committed transactions.
C. Repeat step 1.
3. Consistency of History files
A. For the entire History file, count the History records and sum their delta values.
B. Verify that this History record count equals the sum of History record count
taken in step 2A plus the number of committed transactions.
C. Verify that the difference between the final History delta sum and the initial
History delta sum equals the difference between the final and initial Branch
record balances.

2.3 Isolation Tests
Operations of concurrent transactions must yield results which are indistinguishable

TPC Benchmark B Full Disclosure 5

Clause 2 Related Items

from the results which would be obtained by forcing each transaction to be serially
executed to completion in some order.
The following tests were performed and verified the isolation property of the transactions for conventional locking used by the database system:
Isolation of Completed Transactions
1. Select the Branch balance for a Branch record (Branch B).
2. Start transaction 1.
A. Update the Branch B record with delta(1).
B. Stop just prior to committing transaction 1.
3. Start transaction 2.
A. Attempt to update Branch B with delta(2).
B. Transaction 2 hangs.
4. Resume transaction 1.
A. Update the Teller record.
B. Update the Account record.
C. Insert the History record.
D. Commit transaction 1.
5. Resume transaction 2.
A. Update the Teller record.
B. Update the Account record.
C. Insert the History record.
D. Commit transaction 2.
6. Select the Branch balance for Branch B. The balance should equal the previous
balance plus delta(1) and delta(2).
Isolation of Aborted Transactions
1. Start transaction 1.
A. Update Branch B with delta(3).
B. Stop just prior to committing transaction 1.
2. Start transaction 2.
A. Attempt to update Branch B with delta(4).
B. Transaction 2 hangs.
3. Resume transaction 1.
A. Update the Teller record.
B. Update the Account record.

6 TPC Benchmark B Full Disclosure

Clause 2 Related Items

C. Insert the History record.
D. Abort transaction 1 and perform a rollback recovery.
4. Resume transaction 2.
A. Update the Teller record.
B. Update the Account record.
C. Insert the History record.
D. Commit transaction 2.
5. Select the Branch balance for Branch B. The balance should equal the previous
balance plus delta(4).
The preceding isolation tests were repeated for Teller and Account files (that is, by
generating a conflict on a Teller and then an Account record).

2.4 Durability Tests
The tested system must guarantee the ability to preserve the effects of committed transactions and ensure database consistency after recovery from any one of the failures
listed below:
•

Permanent irrecoverable failure of any single durable medium containing database, ABTH (Accounts, Branch, Teller, and History) files/tables, or recovery log
data.

•

Instantaneous interruption (system crash/system hang) in processing which requires system reboot to recover.

•

Failure of all or part of memory (loss of contents).

The following test was performed for each of the preceding types of failures to verify
the durability property of the SUT:
•

For the entire History file, count the History records.

•

Perform the standard TPC Benchmark B test and record the committed transactions in the success file.

•

Cause one of the preceding types of failure.

•

Restart the system under test for this failure as required in Clause 2.5.3.

•

Verify that every record in the success file has a corresponding record in the History file.

•

For the entire History file,count the History records. Verify that the number of
records in the History file is greater than or equal to the original count obtained
in step 1 plus the number of records in the success file. If they are different, the
new History file must contain additional records and the difference must be less
than or equal to the number of terminals (tellers) simulated.

TPC Benchmark B Full Disclosure 7

Clause 3 Related Items

The durability tests were run on a database sized at 10% of the fully loaded database,
i.e., 5 TPS, in order to comply with future, expected TPC requirements.
In addition, the durability test "failure of all or part of memory (loss of contents),"
i.e., complete power failure, was applied to the fully loaded database under full load
conditions. This too was done to comply with future, expected TPC requirements.

In addition, the test sponsors must guarantee that, to the best of their knowledge, a
fully-loaded system would pass the durability tests.
To the best of the test sponsor’s knowledge, a fully-loaded and fully-scaled system
would pass the durability tests.

3 - Clause 3 Related Items
3.1 ABTH Data Storage Distribution
The distribution across storage media of ABTH (Accounts, Branch, Teller, and History) files/tables and all logs must be explicitly depicted.
The following diagram shows how the databases were distributed on disk media on
the DECsystem 5500 test system for both the benchmark and priced system configurations.
The physical log was placed on a RF31 (381 Mbyte) disk drive using a partition of 350
Mbytes. The rootdbs space containing the logical logs and catalog were located on
another RF31 (381 Mbytes) disk drive using 350 Mbytes for the partition. The
rootdbs mirror (logical log) was located on another RF31.

8 TPC Benchmark B Full Disclosure

Clause 3 Related Items

ABTH Data Storage Distribution Diagram
Kbytes
Used (000’s)

Partition File/ Data

Percent of
Data

RF31

ra0a
ra0b
ra0g

/root
swap dump
/usr

15.6
83
273.4

RF31

rra1c

physical log

350

100.0%

RF31

rra2c

rootdbs (logical logs)

350

100.0%

RF31

rra3c

mirror rootdbs (logical logs)

350

100.0%

RF31

rra4c

account
history

90
282.2

16.6%
13.1%

RF31

rra5c

account
history

90
282.2

16.6%
13.1%

RF31

rra6c

account
history

90
282.2

16.6%
13.1%

RZ56

rrz2g
rrz2h

account
teller
branch
history (active)

90
0.0047
0.047
500

16.6%
100.0%
100.0%
23.3%

RZ56

rrz3c

account
history

90
400

16.6%
18.6%

RZ56

rrz4c

account
history

90
400

16.6%
18.6%

RZ57

rrz0c

logical log archive

801

50.0%

RZ57

rrz1c

logical log archive

801

50.0%

The distribution of the database is further evidenced and illustrated by the Informix
tbstat utility tbstat_d.

TPC Benchmark B Full Disclosure 9

Clause 3 Related Items

tbstat_d Listing

RSAM Version 4.10.UE1P1 -- On-Line -- Up 00:54:59 -- 12016 Kbytes
Dbspaces
address
number
80b47c
1
80b4ac
2
80b4dc
3
80b50c
4
4 active, 20 total

flags
2
1
1
1

fchunk nchunks flags
1
1
M
2
1
N
3
1
N
4
6
N

Chunks
address
chk/dbs
8084fc
1 1
809cbc
1 1
808594
2 2
80862c
3 3
8086c4
4 4
80875c
5 4
8087f4
6 4
80888c
7 4
808924
8 4
8089bc
9 4
9 active, 40 total

offset
500
500
500
500
62500
62500
62500
25000
125000
125000

size
free
175000 163978
175000 0
175000 23242
105000 101222
45000 42
45000 497
45000 497
45000 497
45000 497
45000 997

owner
informix
informix
informix
informix

bpages

flags
POMOPOPOPOPOPOPOPOPO-

name
rootdbs
physdbs
tbhdbs
acctdbs

pathname
/dev/rra2c
/dev/rra3c
/dev/rra1c
/dev/rrz2h
/dev/rra4c
/dev/rra5c
/dev/rra6c
/dev/rrz2g
/dev/rrz3c
/dev/rrz4c

3.1.1 History Storage and Recovery
Within the priced system, there must be sufficient on-line storage to support 8 hours of
recovery log data, if required to recover from any single point of failure, plus any other
expanding system files (see Clause 7.1) and durable history records/rows for 30 eighthour days at the published tpsB rate (i.e., 30 x 8 x 60 x 60 = 864,000 records/rows per
tpsB).
The history and log file storage calculations are shown below:
History File Storage
The following calculations were used to determine the aggregate size of the history
file.
INFORMIX-OnLine Page Size
2048 bytes
Overhead per Page
32 bytes
Overhead per Row
4 bytes
History Table Row Size
50 bytes
History Rows per Page = (Page Size - Page Overhead)/(Row Size + Row Overhead)
truncated to next lowest integer value = 37 History Rows per Page
History Rows Needed = (tpsB * 3600 * 8 * 30) = 38,880,000 Rows
History Space = (Rows Needed/37) * 2048 = 2,101,621.62 Kbytes

10 TPC Benchmark B Full Disclosure

Clause 3 Related Items

Logfile Storage
During the benchmark run, the Informix logical logs were mirrored. In addition, the
inactive logfile segments were archived to disk using INFORMIX-OnLine Continuous
Archiving. In all cases unbuffered logging was used. Two disk drives were used; one
for the logical logs and one for the mirror.
The Informix tbstat utility (tbstat_l) was used to record write data and logfile data
production rates. In the audited reported run, the values were
Number of Writes
Pages/Write

33,675
1.6

The run had a two minute (120 seconds) ramp-up and a 32 minute measurement window. Although the number of writes occurred over the entire 34 minute period, only
the steady state portion of the interval should be used for calculation because during
ramp-up the log write rate would have been less. As a result, logfile space needed
was as follows:
Total logfile storage required/8 hours=
33,675 writes/32 minutes * 1.6 pages/write * 2048 bytes/page =
3,448,320 bytes/minute * 480 minutes/8 hours =
1,655,193,600 bytes/8 hours
Total Logfile Space Needed:
Active Log Space Supplied

1,655,193,600 bytes
- 15,000,000 bytes
-----------------1,640,193,600 bytes

Additional 8 hour log space was required. Two 1 Gbyte drives were used to accommodate this requirement.
In addition, because INFORMIX-OnLine records a timestamp for every completed
logical log archived, we used the timestamp to calculate the average time to archive
one logical log during the steady state run. The average time to fill a 5 Mbyte logical
log was approximately 93 seconds which equates to an 8 hour logfile requirement of
1,548,387,097 bytes.
Because the earlier calculation showed a worst case condition, we used those figures.
We supplied 2,030,000,000 bytes for logical log and archive.
Informix tbstat output for the logical logs and part of the message log follow.

TPC Benchmark B Full Disclosure 11

Clause 3 Related Items

tbstat_l listing
RSAM Version 4.10.UE1P1 -- On-Line -- Up 00:54:59 -- 12016 Kbytes
Physical Logging
Buffer
bufused
P-1
12
phybegin
2006de

bufsize
16
physize
150000

Logical Logging Buffer
bufused
bufsize numrecs
L-3 0
16
749244
address
884408
884424
884440

number
1
2
3

flags
U-B---L
U---C-F------

numpages
84698
phypos
32814
numpages
52766

uniqid
1018
1019
0

numwrits
5297
phyused
4524

numwrits
33675

begin
100fb2
101976
10233a

size
2500
2500
2500

pages/io
15.99
%used
3.02
recs/page
14.2
used
2500
267
0

pages/io
1.6
%used
100.00
10.68
0.00

Message Log File Listing

RSAM Version 4.10.UE1P1 -- On-Line -- Up 00:54:59 -- 12016 Kbytes
Message Log File: /usr/informix/online.log
12:31:39
Logical Log 1011 Complete
12:31:47
Checkpoint Completed
12:32:17
Logical Log 1011 Backed Up
12:33:17
Logical Log 1012 Complete
12:33:34
Logical Log 1012 Backed Up
12:34:51
Logical Log 1013 Complete
12:34:59
Checkpoint Completed
12:35:21
Logical Log 1013 Backed Up
12:36:29
Logical Log 1014 Complete
12:37:08
Logical Log 1014 Backed Up
12:38:03
Logical Log 1015 Complete
12:38:20
Checkpoint Completed
12:38:24
Logical Log 1015 Backed Up
12:39:45
Logical Log 1016 Complete
12:40:10
Logical Log 1016 Backed Up
12:41:19
Logical Log 1017 Complete
12:41:28
Checkpoint Completed
12:41:58
Logical Log 1017 Backed Up
12:42:57
Logical Log 1018 Complete
12:43:14
Logical Log 1018 Backed Up

Appendix E contains a complete listing of the disk devices to support the test.

12 TPC Benchmark B Full Disclosure

Clause 4 Related Items

3.2 Database Contents and Method of Population
A description of how the database was populated, along with sample contents of each
ABTH file/table to meet the requirements described in Clause 3.
Database Contents
Appendix C contains the database population program and Appendix D contains
samples of the contents of the database files used in the tests.

3.3 Type of Database
A statement of the type of database utilized, e.g., relational, Codasyl, flat file, etc.
These TPC Benchmark B tests used INFORMIX-OnLine, a relational database management system.

4 - Clause 4 Related Items
There are no Clause 4 Related Items in the checklist for TPC-B.

5 - Clause 5 Related Items
5.1 Method of Verification of Random Number Generator
The method of verification of the random number generator should be described.
Branch, Teller, and Account IDs were generated by the random number generation
routines, random() and srandom() in the bench.h code. Random()/srandom() use a
non-linear additive feedback random number generator, employing a default table
size of 31 long integers to return successive random numbers in the range from 0 to
(2**31)-1. These routines produce a more random sequence than earlier subroutines
such as rand(). Random() and srandom() are well known random number generation
routines. Randomness of the generated values are further verified by observing the
85/15 distribution rule, which showed that approximately 85% of the transactions
submitted to a Branch had the Account belong to that Branch.

5.2 Horizontal Partitioning Disclosure
Vendors must clearly disclose if horizontal partitioning is used. Specifically, vendors
must:
•

Describe textually the extent of transparency of the implementation

•

Which tables/files were accessed using partitioning

•

How partitioned tables/files were accessed

Horizontal partitioning of the database was not used. Horizontal partitioning, i.e.
the partitioning of a table according to some logical order, was not used. The account
relation records were randomly distributed over multiple (6) disk drives.

TPC Benchmark B Full Disclosure 13

Clause 6 Related Items

6 - Clause 6 Related Items
Report all the data specified in Clause 6.6, including maximum and average residence
time, as well as performance curves for number of transactions vs. residence time (see
Clause 6.6.1) and throughput vs. level of concurrency for three data points (see Clause
6.6.5).
The graphs and tables in this section show the response time performance results.
Please note that for all performance runs the database was scaled for 45 TPS.

14 TPC Benchmark B Full Disclosure

Clause 6 Related Items

Residence Time Frequency Distribution for All Transactions

Number of
Transactions
30000
Average Residence Time: 0.42 seconds

25000
20000
15000

Percentile Residence Time: 0.72 seconds
10000
5000
Maximum Residency Time: 18.45 seconds

0
0.00

1.00

2.00
3.00
Residence Time (seconds)

4.00

5.00

Throughput Versus Level of Concurrency

TPS
50
40
30
20
10
0

CL
0

CR CH

9 12 15 18 21
Level of Concurrency

TPC Benchmark B Full Disclosure 15

Clause 7 Related Items

Concurrency Legend
Measured Points

Level of
Concurrency

TPS

Average Residence
Time (seconds)

40.5

0.35

40.6

0.42

40.3

0.49

Profile of Executed Transactions
Description

Result

Remote Transactions (see Clause 6.6.2)

15.00%

Home Transactions

85.00%

Transactions started and not completed during measurement interval (see Clause 6.6.3)

0.02%

Number of transaction started but not completed

Total number of transactions

77,967

Average residence time for all transactions

0.42 seconds

Maximum residence time for all transactions

18.45 seconds

Percent of all transactions qualified within 2 second
response time constraint

99.81

Maximum qualified throughput

40.6 tpsB

7 - Clause 7 Related Items
7.1 Determining Steady State
The method used to determine that the SUT had reached a steady state prior to commencing the measurement interval should be described.
Confirmation that the SUT has reached steady state prior to the beginning of the
data collection measurement interval is based on a visual inspection of the plot of
TPS versus time.
The design of the benchmark driver program was such that all processes wait to be
signaled to commence ramp-up work. During ramp-up, the processes begin executing identical TPC-B transactions as they do during the steady state run.
During the ramp-up, which lasted for 120 seconds, all processes began executing the

16 TPC Benchmark B Full Disclosure

Clause 7 Related Items

identical TPC-B transaction that they do during the timed steady state run. At the
end of the ramp-up period, each process independently kept track of the numbers and
characteristics of its committed transactions that started during the steady state interval. The audited benchmark steady state period lasted for 32 minutes. When the
run was completed, the processes individually and independently reported their accumulated transactions and residence time results. The driver then calculated the required numbers to report.
To confirm that steady state was reached, the history table was examined. The
graph titled "TPS Versus Time" indicates the number of transactions completed in
each 10 second interval. The steady state portion is labeled on the graph. Note the
pronounced dips (checkpoints) in transaction rate that occurred 10 times during the
steady state for the run. Also, note the less pronounced dip that occurred as each 5
Mbyte logical log is backed up.

TPS Versus Time

TPS

(10 second
intervals)

60
Steady State (32 minutes)

50
40
30
20
10
0
0

15
20
Time (minutes)

7.2 Work Performed During Steady State
A description of how the work normally performed during a sustained test (for example, checkpointing, writing redo/undo log records, etc., as required by Clause 7.2), actually occurred during the measurement interval.
When INFORMIX-OnLine receives a SQL statement from the application, it determines how to best access the data. Using an index (B-tree), INFORMIX-OnLine determines the page number from the database that the record is located on, and
searches for the page in shared memory.
If the page is not in shared memory, INFORMIX-OnLine chooses a LRU Buffer in
shared memory and reads the page from the database into the buffer. Typically this

TPC Benchmark B Full Disclosure 17

Clause 7 Related Items

will take two disk reads. The first read acquires the bottom level of the B-tree index,
the second disk read actually acquires the data.
When a transaction starts, a BEGIN WORK is written to the logical log buffers.
When the application issues a SQL UPDATE statement (Account, Teller, and
Branch) to modify a record, the copy of the record, if it is already in shared memory,
is locked and updated. A transaction record is written to the logical log buffer.
At the same time, if the page in shared memory has not previously been written to, a
copy of the before image of the page is written to the physical log buffer in shared
memory. In addition, the before and after images of the record are written to the
logical log buffer in shared memory. So, the physical log buffer contains a copy of the
page that a record is on, as it looked prior to making any modification.
When the application issues a SQL COMMIT WORK statement, the logical log buffer
is flushed from shared memory to the logical log on disk in a single I/O. The database pages remain in shared memory and are not written to the database at that
time. Any locks that were placed by the transaction are released. This means that
when an application commits a transaction to the database, the logical log buffer is
written to a corresponding logical log on disk with a single I/O, and successful completion code is returned to the application.
Periodically INFORMIX-OnLine will automatically write all modified pages in
shared memory to their appropriate locations in the database during a checkpoint. A
checkpoint is preceded by a write of the physical log buffer to the physical log on disk.
Checkpoints occur periodically during the run. With INFORMIX-OnLine there are
several ways of controlling when a checkpoint occurs. For our benchmark,
checkpointing occurs every time INFORMIX-OnLine starts the last logical log. We
configured INFORMIX-OnLine with three logical logs. Thus, every time two logs
were filled and the third started, a checkpoint would occur. In our benchmark run of
32 minutes, 10 checkpoints occurred.
When the checkpoint occurs, one or more background processes called page cleaners
"wake up" and write all the modified pages from shared memory to the database on
disk. A checkpoint record is written to the logical log buffer. A checkpoint message
is written to the message log.
The page reading and writing activity to the individual chunk partitions in the database are reflected in the Informix utility tbstat_D and tbstat_p.
tbstat_D
RSAM Version 4.10.UE1P1 -- On-Line -- Up 00:54:59 -- 12016 Kbytes
Dbspaces
address
number
80b47c
1
80b4ac
2
80b4dc
3
80b50c
4
4 active, 20 total

flags
2
1
1
1

fchunk
1
2
3
4

18 TPC Benchmark B Full Disclosure

nchunks flags
1
M
1
N
1
N
6
N

owner
informix
informix
informix
informix

name
rootdbs
physdbs
tbhdbs
acctdbs

Clause 7 Related Items

Chunks
address
chk/dbs
8084fc
1
1
809cbc
1
1
808594
2
2
80862c
3
3
8086c4
4
4
80875c
5
4
8087f4
6
4
80888c
7
4
808924
8
4
8089bc
9
4
9 active, 40 total

offset
500
500
500
500
62500
62500
62500
25000
125000
125000

page Rd
52503
0
0
4099
27850
27550
28227
28354
27896
27512

page Wr
52784
52784
84693
2729
13261
13163
13486
13543
13295
13112

pathname
/dev/rra2c
/dev/rra3c
/dev/rra1c
/dev/rrz2h
/dev/rra4c
/dev/rra5c
/dev/rra6c
/dev/rrz2g
/dev/rrz3c
/dev/rrz4c

tbstat_p

RSAM Version 4.10.UE1P1 -- On-Line -- Up 00:54:59 -- 12016 Kbytes
Profile
dskreads pagreads bufreads %cached dskwrits pagwrits bufwrits %cached
174482 223973 2416938 92.78 154138 272843 337867 54.38
isamtot open start read
write rewrite delete commit rollbk
3071705 97 249692 250137 83274 249633 0
83517 0
ovtbls ovlock ovuser ovbuff usercpu syscpu numckpts flushes
0
0
0
0
951.87 682.60 10
20
bufwaits lokwaits lockreqs deadlks dltouts lchwaits ckpwaits compress
6892
10362 1946418 0
0
52552 149
1

7.3 Determining Reproducibility
A description of the method used to determine the reproducibility of the measurement
results.
Experiments were repeated at least 3 times at the maximum targeted TPS level to
ensure reproducibility. The results are shown in the following table. The variation
in TPS was less than 2%.

TPC Benchmark B Full Disclosure 19

Clause 8 Related Items

DECsystem 5500 TPC-B Benchmark Runs
Percent
Run #

1
2
3

Processes CPUs tpsB

15
15
15

1
1
1

39.9
40.6
40.5

< 2 sec.

99.78
99.81
99.81

db
Transactions

76,627
77,967
77,831

Size

Duration

45 tps
45 tps
45 tps

32.0 mins
32.0 mins
32.0 mins

7.4 Duration of Measurement Period
A statement of the duration of the measurement period for the reported tpsB (it should
be at least 15 minutes and no longer than 1 hour).
Each experiment used a measurement period of 32 minutes and began approximately
2 minutes after all servers had begun executing transactions.

8 - Clause 8 Related Items
8.1 Description of the Driver
If the driver is commercially available, then its inputs should be specified. Otherwise,
a description of the driver should be supplied.
The driver used was an "internal driver" (i.e., the driver software resides on the system under test, not on a remote driver machine) that controls transaction processing
and performance data collection for the TPC Benchmark B runs. The driver was
comprised of two parts: a control csh script and a set of identical ESQL/C transaction
programs that submitted the TPC Benchmark B transactions for execution.
The control script performs the following operations:
1. forks and execs the desired number of transaction programs, passing ramp-up
and measurement interval parameters as command line arguments.
2. waits for a short period of time (30 seconds) to ensure that each driver has
started up and opened the test database.
3. sends a SIGUSR1 signal to each transaction process to synchronize the start of
transaction processing.
4. waits until all transaction processes have completed the benchmark run.
5. invokes a program called sumrun to sum the performance statistics collected by
the transaction processes involved in the benchmark run.
After each transaction program completes a benchmark run, the transaction program stores residence time counts, incomplete transaction counts, and other performance statistics in a database table named "results". The sumrun program

20 TPC Benchmark B Full Disclosure

Clause 9 Related Items

sums all "results" records for a run and inserts aggregate run values into a table
named "runs".
Each transaction program performs the following operations:
1. examines its command line arguments to determine the ramp-up and measurement intervals to use.
2. waits until it receives a SIGUSR1 signal before initiating transaction processing.
3. continuously submits TPC-B transactions, with 0 sleep time. The transaction
program collects response time statistics in internal program data structures, but
does not begin collecting them until the ramp-up period has completed.
4. inserts its collected performance statistics into a "results" table record once the
measurement interval has completed. It is the contents of these "results" records
that are summed by the sumrun program.
"Success files" were implemented through the tpc.ec application program by writing
synchronously using fsync() and flushing the confirmation of transactions to standard output. This was captured into a file nohup.out running under the Korn shell.

9 - Clause 9 Related Items
9.1 Hardware and Software Components
A detailed list of hardware and software used in the priced system. Each item must
have vendor part number, description, and release/revision level, and either general
availability status or committed delivery date. If package pricing is used, contents of
the package must be disclosed.

9.1.1 Priced System Configuration Tables
The following tables show the hardware and software components in the priced
DECsystem 5500 system:

TPC Benchmark B Full Disclosure 21

Clause 9 Related Items

Component

Product

Processor

DECsystem 5500

Quantity
1
32 Megabytes

Memory
TK50
DSSI, SCSI

Disk controller
Disks

RF31

RZ57

Tape drive

Operating system
Database

1 of each

RZ56

ULTRIX 4.2
INFORMIX-OnLine V4.10
INFORMIX ESQL/C

9.1.2 Package Pricing
Package Description
---------------------------

Model #
-----------

DECsystem 5500 120/240V
BA430 Pedestal Enclosure
32 MB Memory
SCSI, DSSI Storage Adapter
ThinWire/ThickWire Ethernet
Included Software Licenses
ULTRIX 4 User License
UWS Server Support License
Prestoserve  License
English Language H/W Doc
US 120 V Power Cord
With One Year System Warranty

DU-55HT1-A9

Package Description
---------------------------

Model #
-----------

2xRZ57 1GB Disks with
BA42 Box

SZ12C-CA

Package Description
---------------------------

Model #
-----------

TK50 Tape Drive with
Tape Controller and Box

TK50Z-GA

Package Description
--------------------------2xRZ56 665 MB Disks with
BA42 box

Model #
----------SZ12B-BA

Package Description
--------------------------5xRF31 w/R400X in Exp. Cab.

Model #
-----------DL-RF31A-A5

22 TPC Benchmark B Full Disclosure

Clause 9 Related Items

9.2 Total Price of System Configuration
The total price of the entire configuration is required including: hardware, software
and maintenance charges. Separate component pricing is recommended. The basis of
all discounts used shall be disclosed.
This section lists the separate components in the priced system and their associated
purchase and maintenance costs. All items are currently available. All prices were
taken from the Digital Standard Pricing System (DSPS) on November 21, 1991. A
description of the packages used in the pricing is contained in Section 9.1.2.
Informix prices were taken from Informix price list, titled "Americas Price List, Advance Products, Release 4.0 or Greater, Class D", dated August 1, 1991.

9.2.1 Hardware Pricing
The Digital TPC Benchmark B DECsystem 5500 test used packaged hardware systems whenever possible to simplify configurations to the fewest number of line items.
Disks were connected using DSSI and SCSI controllers. The system used a TK50
tape drive to load the software and back up the database.
The purchase price of all systems includes one year of hardware warranty service.
Post-warranty hardware service is configured for an additional four years.
The following levels of post-warranty hardware service are used in the system pricing:
DECsystem Support 9x5 (DS9) and 2-4 hours response time.
Basic Monthly Charge (BMC) warranty level is the same as the DS9 to which the
hardware is directly attached.
Basic System Support (BSS) with a warranty upgrade to DS9.

9.2.2 Software Pricing
The priced system uses the following software products:
•

ULTRIX V4.2 operating system

•

INFORMIX-OnLine relational database management system

•

INFORMIX-ESQL/C

The ULTRIX license purchase includes one year of warranty service. Post-warranty
service is configured for an additional four years. The software warranty and service
level are the same as the service level for the hardware system on which the software
operates. The level of post-warranty service is Software Support Service (SSS).

9.2.3 Price Discounts
Digital’s five (5) years warranty pricing is as follows:

TPC Benchmark B Full Disclosure 23

Clause 9 Related Items

•

the unit price carries one (1) year warranty.

•

the price of year 2-5 warranty adder is calculated according to this formula:
-

(warranty/month)*12*(1+1+1.07+(1.07)²)=(1.053725*48*(warranty/month))

The pre-payment maintenance (warranty) discount is calculated at 25% of the year
2-5 warranty price.
Informix’s five-year prepaid maintenance option consists of five years of maintenance
for four times the price of standard maintenance.

24 TPC Benchmark B Full Disclosure

9.2.4 System Pricing Summary
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

DECsystem TPC-B = 40.6 TPS
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------US LIST
UNIT PRICE
TOTAL
SERVICE
MAIN.
2-5 YRS
PRICE+SRVC
DESCRIPTION
MODEL #
1 YR WARR
QTY
PRICE
LEVEL
$/MO.
# MO MAIN. PRICE
5 YR COST
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------Digital Price (21 November 1991)
-------------Host and Database
----------------------DS5500 BA430, 32 MB
Warranty Upgrade To DS9
2 RZ56 665 MB Disk in BA42 Box
1 RZ56 665 MB Disk in BA42 Box
2 RZ57 1.0 GB Disk in Exp. Box
5xRF31 w/R400X Exp. Cab.
RF31 381 MB Disk
TK50 Tape Controller Box VJ3100
ULTRIX-32 V4.2 Media & Doc.

DU-55HT1-A9
FM-DECUP-12
SZ12B-BA
SZ12B-XA
SZ12C-CA
DL-RF31A-A5
RF31E-AA
TK50Z-GA
QA-VYVAA-H5

$38,340.00
$648.00
$9,740.00
$5,120.00
$16,324.00
$23,900.00
$4,800.00
$3,860.00
$3,240.00
‘

1
1
1
1
1
1
2
1
1

Digital Subtotal
Years 2-5 Warranty Adder =5.3725%

$38,340.00
$648.00
$9.740.00
$5,120.00
$16,324.00
$23,900.00
$9,600.00
$3,860.00
$3,240.00
----------------$110,772.00

BSS
DS9
DS9/BMC
DS9/BMC
DS9/BMC
DS9/BMC
DS9/BMC
DS9/BMC
NA

$0.00
$549.00
$120.00
$60.00
$152.00
$152.00
$25.00
$30.00
$.00

48
48
48
48
48
48
48
48
48

$/YEAR
----------

YEAR
---------

$590.00
$200.00

4
4

----------------$110,772.00

Digital Subtotal
Prepayment Maintenance Discount=25%

----------------$110,772.00

Digital Total

$0.00
$26,352.00
$5,760.00
$2,880.00
$7,296.00
$7,296.00
$2,400.00
$1,440.00
$.00
--------------$53,424.00
$2,870.20
--------------$56,294.20
($14,073.55)
--------------$42,220.65

$38,340.00
$27,000.00
$15,500.00
$8,000.00
$23,620.00
$31,196.00
$12,000.00
$5,300.00
$3,240.00
--------------------$164,196.00
$2,870.20
--------------------$167,066.20
($14,073.55)
--------------------$152,992.65

$2,360.00
$800.00
---------------$3,160.00

$5,660.00
$1,460.00
---------------------$7,120.00

$45,380.65

$160,112.65

Informix Price (1 August 1991)

INFORMIX-OnLine (Class "D" License) Full Dev./Run T
INFORMIX-ESQL/C
Full Dev./Run T

$3,300.00
$660.00

1
1

Informix Total

$3,300.00
$660.00
---------------$3,960.00

CONFIGURATION TOTALS

$114,732.00

SSS
SSS

TPS
$/TPS

40.6
$3,944

Clause 10 Related Items

9.3 Performance and Price/Performance
A statement of the measured tpsB, and the calculated price/tpsB.
The following table shows measured tpsB and price/tpsB results for the tested
system:

TPS

Price per TPS

CPU Model

Software

(tpsB)

($/tpsB)

DECsystem 5500

ULTRIX 4.2 and
INFORMIX-OnLine
4.10

40.6

$3,944

10 - Clause 10 Related Items
None.
11 - Clause 11 Related Items
11.1 Independent Auditor’s Report
If the benchmark has been independently audited, then the auditor’s name, address, phone number, and a brief audit summary report indicating compliance
must be included in the full disclosure report. A statement should be included,
specifying when the complete audit report will become available and who to contact in order to obtain a copy.
Appendix G contains the complete independent auditor’s report by KPMG Peat
Marwick for the tests described in this report.

26 TPC Benchmark B Full Disclosure

Appendix A
Application Code
This appendix contains the source code of the application programs that implement
the TPC Benchmark B transaction.

A.1 tpc.ec source code
#include <stdio.h>
#include <sys/signal.h>
#include <sys/types.h>
#include <sys/timeb.h>
#include <math.h>
$include sqlca ;
#include "bench.h"
$long acct_bal, cntr, seconds, intvl, startsec, tot_response ;
$int branch_num, teller_num, acct_num, delta, acct_branch, run, procnum ;
$int notdone, tmslot[BUCKETS+1] ;
int rampup, runtime, timing, thru, measure, bucketval, transactions, verbose ;
int longest_tran;
settimer() { timing = ~timing ; }
setmeasure()
{
intvl = (measure) ? rampup : runtime ;
thru++ ;
measure = ~measure ;
startsec = time(0) ;
}
main(argc,argv)
int argc ;
char **argv ;
{
int i, *rnum, do_trans() ;
runtime = rampup = intvl = 0 ;
transactions = -1 ;
procnum = atoi(argv[1]) ;
i=1;
while (++i < argc) {
if (strcmp(argv[i], "-s") == 0)
runtime += atoi(argv[++i]) ;
else if (strcmp(argv[i], "-m") == 0)
runtime += (60 * atoi(argv[++i])) ;
else if (strcmp(argv[i], "-h") == 0)
runtime += (3600 * atoi(argv[++i])) ;

TPC Benchmark B Full Disclosure A-1

Appendix A Application Code

else if (strcmp(argv[i], "-t") == 0)
transactions = atoi(argv[++i]) ;
else if (strcmp(argv[i], "-r") == 0)
rampup = atoi(argv[++i]) ;
else if (strcmp(argv[i], "-v") == 0)
verbose = 1 ;
else {
fprintf(stderr,"usage: tp1 <proc #> [-t <n>] [-r <n>]
[-h <n> -m <n> -s <n>]\n") ;
exit(1) ;
}
}
RandSeed(getpid()) ;
if (runtime == 0)
runtime = (transactions == -1) ? 300 : 30000 ;
printf("process %d: procnum=%03d runtime=%d seconds / %d transac
tions\n", getpid(),procnum,runtime,transactions) ;
cntr = tot_response = run = measure = timing = thru = intvl = notdone = 0 ;
for (i=0; i < 50; i++)
tmslot[i] = 0 ;
$ database tpc ;
SqlErr("attach to database") ;
$ select max(number) into $run from results ;
SqlErr("select from results") ;
if (run < 0)
run = 0 ;
++run ;
do_prepare() ;
bucketval = RPTINTVL * 1000 / BUCKETS ;
intvl = rampup ;
signal(SIGUSR1,settimer) ;
sigpause(0) ;
printf("%d starting\n",procnum) ; do_trans() ; testend() ;
}
do_prepare()
{
$char s[512];
sprintf(s,"%s %s %s%d%s %s commit work",
"update account set balance = balance + ? where current of sel_acct;",
"update teller set balance = balance + ? where number = ?;",
"insert into history",procnum % HISTORY,
" values(?,?,?,?,CURRENT YEAR TO SECOND,’the rest is history’);",
"update branch set balance = balance + ? where number = ?;") ;

A-2 TPC Benchmark A Full Disclosure

Appendix A

Application Code

$ prepare tpc_trans from $s;
SqlErr("prepare updall");
$ prepare bwork from "begin work" ;
SqlErr("prepare begin work") ;
$ declare sel_acct cursor for
select balance into $acct_bal from account
where number = $acct_num
for update of balance ;
SqlErr("declare cursor") ;
$ set isolation to cursor stability ;
SqlErr("set isolation") ;
$ set lock mode to wait;
SqlErr("set lock mode");
}
do_trans()
{
long timediff ;
char s[100] ;
struct timeb clk_beg,clk_end ;
startsec = time(0) ;
if (rampup == 0)
setmeasure() ;
else
thru++ ;
while (timing && (cntr != transactions)) {
/*
* select a random branch, a random teller at that branch, and
* 85% of the time a random account at that branch, and 15% of
* the time a random acccount at a different branch.
*/
teller_num = RandVal() % T_RECS ;
branch_num = teller_num / T_PERB ;
acct_num = RandVal() % A_PERB ;
if ((RandVal() % 100) < 85)
acct_branch = branch_num ;
else {
do
/* endless loop when TPS_SIZE=1 */
acct_branch = RandVal() % B_RECS ;
while (acct_branch == branch_num) ;
}
acct_num = acct_branch * A_PERB + acct_num ;
delta = RandVal() % 1999999 - 999999 ;
if (measure)
notdone++ ;

TPC Benchmark B Full Disclosure

A-3

Appendix A Application Code

ftime(&clk_beg) ;
$ execute bwork ;
$ open sel_acct ;
SqlErr("open cursor") ;
$ fetch sel_acct ;
SqlErr("fetch cursor") ;
if (sqlca.sqlcode == 0) {
$ execute tpc_trans using
$delta,
$delta, $teller_num,
$acct_num, $teller_num, $branch_num, $delta,
$delta, $branch_num ;
}
ftime(&clk_end) ;
if (sqlca.sqlcode != 0) {
sprintf(s,"in transaction %d acc#: %d branch#: %d teller#: %d",
cntr, acct_num, branch_num, teller_num) ;
SqlFatal(s) ;
/*
$ rollback work ;
*/
}
timediff = clk_end.time - startsec ;
if (timediff > intvl) {
if (thru == 2)
settimer() ;
else
setmeasure() ;
}
if (measure) {
timediff = (clk_end.time - clk_beg.time) * 1000
+ clk_end.millitm - clk_beg.millitm ;
if(timediff > longest_tran)
longest_tran = timediff;
tot_response += timediff ;
timediff /= bucketval ; /* 0-.124, .125-0.249, etc. seconds */
if (timediff > BUCKETS)
timediff = BUCKETS ;
tmslot[timediff]++ ;
cntr++ ;
if(verbose)
{
printf("procnum %3d: tran %d completed!\n",procnum,cntr);
fflush(stdout);
fsync(1);
}

A-4 TPC Benchmark A Full Disclosure

Appendix A

Application Code

notdone-- ;
}
}
seconds = (transactions > 0) ? (time(0)-startsec) : runtime ;
}
testend()
{
int hrs, min, sec ;
hrs = seconds / 3600 ;
min = (seconds - hrs * 3600) / 60 ;
sec = seconds - hrs * 3600 - min * 60 ;
printf("procnum %3d completed %6d transactions in %4d:%02d:%02d, longest=%d msec.\n",
procnum, cntr, hrs, min, sec, longest_tran) ;
$ insert into results values(
$run, $procnum, $seconds, $cntr, $notdone, $tot_response,
$tmslot[0],$tmslot[1],$tmslot[2],$tmslot[3],$tmslot[4],
$tmslot[5],$tmslot[6],$tmslot[7],$tmslot[8],$tmslot[9],
$tmslot[10],$tmslot[11],$tmslot[12],$tmslot[13],$tmslot[14],
$tmslot[15],$tmslot[16],$tmslot[17],$tmslot[18],$tmslot[19],
$tmslot[20],$tmslot[21],$tmslot[22],$tmslot[23],$tmslot[24],
$tmslot[25],$tmslot[26],$tmslot[27],$tmslot[28],$tmslot[29],
$tmslot[30],$tmslot[31],$tmslot[32],$tmslot[33],$tmslot[34],
$tmslot[35],$tmslot[36],$tmslot[37],$tmslot[38],$tmslot[39],
$tmslot[40]) ;
SqlErr("insert into results") ;
$ close database ;
SqlErr("close database") ;
}

TPC Benchmark B Full Disclosure

A-5

Appendix A Application Code

A.2 createdb.ec source code
#include <stdio.h>
#include "bench.h"
$include sqlca ;
/*
* FILE: createdb.ec (for OnLine)
*
* Creates the database and related tables, except result-consolidation
* tables. It is possible to place the tables on different drives by
* adding location options to the CREATE TABLE statements.
*
* You can also decide to place logging on the database by adding it
* to the CREATE DATABASE statement. However, the loading programs
* provided assume no transaction logging, so you should turn on logging
* afterward via archiving and changing the database logging mode.
*
* The configuration here accommodates scaling to 100 TPS.
*
*/
main()
{
$ create database tpc in TBHDBS ;
SqlErr("create database") ;
$ grant dba to public ;
SqlErr("grant dba") ;
printf("Database created, permission granted\n") ;
$ create table branch
(
number numeric(2,0),
balance numeric(10,0),
fillstr char(92)
)
lock mode row
;
SqlErr("create branch") ;
printf("Branch created\n") ;
$ create table teller
(
number numeric(4,0),
balance numeric(10,0),
branch numeric(2,0),
fillstr char(89)
)
extent size 200
next size 100
lock mode row

A-6 TPC Benchmark A Full Disclosure

Appendix A

Application Code

;
SqlErr("create teller") ;
printf("Teller created\n") ;
$ create table account
(
number
numeric(8,0),
balance
numeric(10,0),
branch
numeric(2,0),
fillstr
char(87)
)
in acctdbs
extent size 5000
next size 1000
;
SqlErr("create account") ;
printf("Account created\n") ;
$ close database ;
SqlErr("close database") ;
exit(0) ;
}

TPC Benchmark B Full Disclosure

A-7

Appendix A Application Code

A.3 createhist.ec source code
#include <stdio.h>
#include "bench.h"
$include sqlca ;
/*
* FILE: createhist.ec (for OnLine)
*
* Creates the history tables. Number of tables is HISTORY in "bench.h".
*
* The configuration here accommodates scaling to 100 TPS.
*
*/
main()
{
$char dstr[200] ;
$int cnt, i ;
$ database tpc ;
SqlErr("connect to database") ;
$ select count(*) into $cnt from systables
where tabname matches "hist" ;
SqlErr("test for history tables") ;
$ select count(*) into $cnt from systables
where tabname matches "hist*" ;
SqlErr("test for history tables") ;
if (cnt) {
printf("Dropping History tables...\n") ;
for (i=0; i < cnt; i++) {
sprintf(dstr,"drop table history%d",i) ;
$ prepare drop_tab from $dstr ;
SqlErr("prepare drop") ;
$ execute drop_tab ;
SqlErr(dstr) ;
}
}
for (i=0; i < HISTORY; i++) {
sprintf(dstr, "%s%d (%s,%s,%s,%s,%s,%s) %s %s %s",
"create table history", i,
"account integer",
"teller integer",
"branch integer",
"delta char(11)",
"tstamp datetime year to second",
"fillstr char(22)",
"extent size 1000",
"next size 1000",

A-8 TPC Benchmark A Full Disclosure

Appendix A

Application Code

"lock mode row"
);
$ prepare make_tab from $dstr ;
SqlErr("prepare create") ;
$ execute make_tab ;
SqlErr("execute history") ;
printf("History%d table created\n",i) ;
}
$ close database ;
SqlErr("close database") ;
exit(0) ;
}

TPC Benchmark B Full Disclosure

A-9

Appendix A Application Code

A.4 createruns.ec source code
#include <stdio.h>
#include "bench.h"
$include sqlca ;
/*
* FILE: createruns.ec
*
* Creates the results tables for cumulative reporting
*
*/
main()
{
$int cnt ;
$ database tpc ;
SqlErr("open database") ;
$ select count(*) into $cnt from systables
where tabname = "runs" ;
SqlErr("test for runs table") ;
if (cnt) {
$ drop table runs ;
SqlErr("drop table runs") ;
}
$ select count(*) into $cnt from systables
where tabname = "results" ;
SqlErr("test for results table") ;
if (cnt) {
$ drop table results ;
SqlErr("drop table results") ;
}
$ create table runs
(
num
numprocs
test_intvl
total_xact
total_inc
resp_time
cpus
test_size
tslot01
tslot02
tslot03
tslot04
tslot05
tslot06

serial,
integer,
integer,
integer,
integer,
integer,
integer,
integer,
integer,
integer,
integer,
integer,
integer,
integer,

A-10 TPC Benchmark A Full Disclosure

Appendix A

Application Code

tslot07
integer,
tslot08
integer,
tslot09
integer,
tslot10
integer,
tslot11
integer,
tslot12
integer,
tslot13
integer,
tslot14
integer,
tslot15
integer,
tslot16
integer,
tslot17
integer,
tslot18
integer,
tslot19
integer,
tslot20
integer,
tslot21
integer,
tslot22
integer,
tslot23
integer,
tslot24
integer,
tslot25
integer,
tslot26
integer,
tslot27
integer,
tslot28
integer,
tslot29
integer,
tslot30
integer,
tslot31
integer,
tslot32
integer,
tslot33
integer,
tslot34
integer,
tslot35
integer,
tslot36
integer,
tslot37
integer,
tslot38
integer,
tslot39
integer,
tslot40
integer,
tslot41
integer
);
SqlErr("create runs") ;
printf("Runs table created\n") ;
$ create table results
number
procnum
seconds
xactcnt
notdone
response
tslot01
tslot02
tslot03
tslot04

(
integer,
integer,
integer,
integer,
integer,
integer,
integer,
integer,
integer,
integer,

TPC Benchmark B Full Disclosure

A-11

Appendix A Application Code

tslot05
integer,
tslot06
integer,
tslot07
integer,
tslot08
integer,
tslot09
integer,
tslot10
integer,
tslot11
integer,
tslot12
integer,
tslot13
integer,
tslot14
integer,
tslot15
integer,
tslot16
integer,
tslot17
integer,
tslot18
integer,
tslot19
integer,
tslot20
integer,
tslot21
integer,
tslot22
integer,
tslot23
integer,
tslot24
integer,
tslot25
integer,
tslot26
integer,
tslot27
integer,
tslot28
integer,
tslot29
integer,
tslot30
integer,
tslot31
integer,
tslot32
integer,
tslot33
integer,
tslot34
integer,
tslot35
integer,
tslot36
integer,
tslot37
integer,
tslot38
integer,
tslot39
integer,
tslot40
integer,
tslot41
integer
);
SqlErr("create results") ;
printf("Results table created\n") ;
$ close database ;
SqlErr("close database") ;
exit(0) ;
}

A-12 TPC Benchmark A Full Disclosure

Appendix A

Application Code

A.5 createidx.ec source code
#include <stdio.h>
#include "bench.h"
$include sqlca ;
/*
* FILE: createidx.ec
*
* Creates the indices for the main database tables. This is a separate
* process in case loads without indices are desired.
*
*/
main()
{
$ database tpc ;
SqlErr("open database") ;
$ create unique index ibranch on branch(number) ;
SqlErr("create branch index") ; printf("Branch index created\n") ;
$ create unique index iteller on teller(number) ;
SqlErr("create teller index") ; printf("Teller index created\n") ;
$ create unique index iaccount on account(number) ;
SqlErr("create account index") ;
printf("Account index created\n") ;
$ close database ;
SqlErr("close database") ;
exit(0) ;
}

TPC Benchmark B Full Disclosure

A-13

Appendix A Application Code

A.6 config.scr source code
echo Going into Quiescent mode
tbmode -uy
echo Creating physdbs...
tbspaces -c -d physdbs -p /dev/rra1c -o 1000 -s 350000
echo Creating tbhdbs...
tbspaces -c -d tbhdbs -p /dev/rrz2h -o 1000 -s 210000
echo Creating acctdbs...
tbspaces -c -d acctdbs -p /dev/rra4c -o 125000 -s 90000
echo Adding chunk to acctdbs...
tbspaces -a acctdbs -p /dev/rra5c -o 125000 -s 90000
echo Adding chunk to acctdbs...
tbspaces -a acctdbs -p /dev/rra6c -o 125000 -s 90000
echo Adding chunk to acctdbs...
tbspaces -a acctdbs -p /dev/rrz2g -o 50000 -s 90000
echo Adding chunk to acctdbs...
tbspaces -a acctdbs -p /dev/rrz3c -o 250000 -s 90000
echo Adding chunk to acctdbs...
tbspaces -a acctdbs -p /dev/rrz4c -o 250000 -s 90000
echo Moving Physical Log
tbparams -p -s 300000 -d physdbs -y
echo Going back On-Line
tbmode -m echo Configuration
done

A-14 TPC Benchmark A Full Disclosure

Appendix A

Application Code

A.7 bench.h code
*
PURPOSE: to set up the sizing of the TPC database
*
* the scale factors for TPC per 1 TPS are:
*
1 Branch, 10 Tellers, 100000 Accounts
*
* Modify the TPS_SIZE to the desired rating.
* DO NOT modify any but the first 4 lines. *
*/
#define TPS_SIZE
#define HISTORY
#define RandVal
#define RandSeed
#define BUCKETS
#define RPTINTVL

45
1
random
srandom
40
5

#define T_PERB
#define A_PERB

10
100000

#define B_RECS
#define T_RECS
#define A_RECS

TPS_SIZE
(T_PERB * B_RECS)
(A_PERB * B_RECS)

#define IsqlCode
#define IsamCode
#define SqlErr(x)
#define SqlErrNF(x)

sqlca.sqlcode
sqlca.sqlerrd[1]
if (IsqlCode) Sqlmsg(x)
if (IsqlCode && IsqlCode != SQLNOTFOUND) Sqlmsg(x)

TPC Benchmark B Full Disclosure

A-15

Appendix B Database Definitions

Appendix B
Database Definitions

#************************************************************************************************
#
#
INFORMIX SOFTWARE, INC.
#
# Title:
tbconfig.std
# Sccsid:
@(#)tbconfig.std 7.2 11/20/90 11:06:55
#Description:
INFORMIX-OnLine Configuration Parameters
#
#************************************************************************************************
# Root Dbspace Configuration
ROOTNAME
ROOTPATH
ROOTOFFSET
ROOTSIZE

rootdbs
/dev/rra2c
1000
350000

# Root dbspace name
# Path for device containing root dbspace
# Offset of root dbspace into device (Kbytes)
# Size of root dbspace (Kbytes)

# Disk Mirroring Configuration Parameters
MIRROR
MIRRORPATH
MIRROROFFSET

1
/dev/rra3c
1000

# Mirroring flag (Yes = 1, No = 0)
# Path for device containing mirrored root
# Offset into mirrored device (Kbytes)

# Physical Log Configuration
PHYSDBS
PHYSFILE

physdbs
300000

# Location (dbspace) of physical log
# Physical log file size (Kbytes)

# Logical Log Configuration
LOGFILES
LOGSIZE

3
5000

# Number of logical log files
# Logical log size (Kbytes)

# Message Files
MSGPATH
CONSOLE

/usr/informix/online.log
/usr/informix/console.log

# System message log file path
# System console message path

# System Archive Tape Device
TAPEDEV
TAPEBLK
TAPESIZE

/dev/null
16
90000

# Tape device path
# Tape block size (Kbytes)
# Maximum amount of data to put on tape
# (Kbytes)

# Log Archive Tape Device
LTAPEDEV
LTAPEBLK

/dev/rrz1c
16

# Log tape device path
# Log tape block size (Bytes)

B-1 TPC Benchmark B Full Disclosure

LTAPESIZE

1000472

# Max amount of data to put on log tape (Kbytes)

0
dectpc
30
0

# Unique id corresponding to an OnLine instance
#
# max time to wait of lock in distributed env.
# Forced residency flag (Yes = 1, No = 0)

# System Configuration
SERVERNUM
SERVERNAME
DEADLOCK_TIMEOUT
RESIDENT

# Shared Memory Parameters
USERS

LOCKS
BUFFERS
TBLSPACES
CHUNKS
DBSPACES
PHYSBUFF
LOGBUFF
LOGSMAX
CLEANERS
SHMBASE
CKPTINTVL

5000
5000
1200
40
20
32
32
3
8
0x800000
720

# Maximum number of concurrent users (proc
# esses)
# Maximum number of locks
# Maximum number of shared buffers
# Maximum number of open tblspaces
# Maximum number of chunks
# Maximum number of dbspaces
# Physical log buffer size (Kbytes)
# Logical log buffer size (Kbytes)
# Maximum number of logical log files
# Number of buffer cleaner processes
# Shared memory base address
# Check point interval (in sec)

2048

# Page size (do not change!)

# System Page Size
BUFFSIZE

#System LRU Parameters
LRUS
LRU_MAX_DIRTY
LRU_MIN_DIRTY
LRU_SEARCH

8
60
50
70

#Number of LRU’s
#Start page cleaning
#Stop page cleaning
#First Level search for free buffers

TPC Benchmark B Full Disclosure B-2

Appendix C Code to Populate Database

Appendix C
Code to Populate Database
This appendix contains the program used to populate the database used in the TPC
Benchmark B tests.

C.1 Database Population Program
The following program was used to populate the database:

#include <stdio.h>
#include <math.h>
#include <sys/types.h>
#include <sys/wait.h>
#include "bench.h"
$include sqlca ;
/*
* FILE: load_db.ec
*
* PURPOSE: load the Branch and Teller tables, and kick off the Account
*
table load procedures. The Account table is loaded by
*
dividing the key range into equal parts (according to the
*
number of load processes), and the "load_act" program is
*
forked off for each process. The program then waits for
*
them to finish and reports the total load time.
*
* NOTE: The type "pid_t" may be system-dependent. Under Ultrix it’s
* equivalent to an "int".
*
*/
FILE *flog,*fopen() ;
int logfile ;
main(argc,argv)
int argc ;
char *argv[] ;
{
int i, load_procs, skip, freespace ;
char begnum[15], endnum[15], log_fname[40], rpt_str[80] ;
long load_accts, startacct, acct_hunk, beg_time, end_time, totsecs ;
pid_t pid ;
union wait wait_status ;
$int branch, teller, branch_idx ;
$char filler[100] ;
RandSeed(getpid()) ;

C-1 TPC Benchmark B Full Disclosure

load_procs = 1 ;
i = logfile = skip = freespace = branch_idx = 0 ;
while (++i < argc) {
if (strcmp(argv[i], "-p") == 0)
load_procs = atoi(argv[++i]) ;
else if (strcmp(argv[i], "-s") == 0)
skip = atoi(argv[++i]) ;
else if (strcmp(argv[i], "-f") == 0)
freespace = atoi(argv[++i]) ;
else if (strcmp(argv[i], "-l") == 0) {
strcpy(log_fname,argv[++i]) ;
logfile = 1 ;
}
else {
printf("usage: load_db -p <#> -s <#> f <#> -l <file>\n") ;
exit(0) ;
}
}
if (load_procs && ((A_PERB % load_procs) != 0)) {
printf("Cannot split up load of accounts evenly. Try
again.\n") ;
exit(0) ;
}
load_accts = (B_RECS - skip) * A_PERB ;
if (load_procs) acct_hunk

load_accts

load_procs ;
for (i=0; i < 10; i++)
bycopy("1234567890",&filler[i*10],10) ;
if (logfile) { if ((flog=fopen(log_fname,"w")) == NULL) {
perror("on opening log file") ;
logfile = 0 ;
}
}
$ database tpc ;
SqlErr("database open") ;
if (freespace) {
$ select count(*) into $branch_idx from sysindexes
where idxname = "ibranch" ;
SqlErr("load_db -- select branch index") ;
if (branch_idx) {
$ drop index ibranch ;
SqlErr("load_db -- delete branch index") ;
}
}
for (branch=skip; branch < B_RECS; branch++) {
$ insert into branch values($branch, 0, $filler) ;

TPC Benchmark B Full Disclosure C-2

Appendix C Code to Populate Database

SqlErr("load_db -- branch insert") ;
for (i=0; i < freespace; i++) {
$ insert into branch values(0, -1, $filler) ;
SqlErr("load_db -- branch free insert") ;
}
}
if (freespace) {
$ delete from branch where balance < 0 ;
SqlErr("load_db -- delete branch records") ;
}
if (branch_idx) {
$ create unique index ibranch on branch(number) ;
SqlErr("load_db -- create branch index") ;
}
print_log("branch table loaded") ;
for (teller=T_PERB*skip; teller < T_RECS; teller++) {
branch = teller / T_PERB ;
$ insert into teller values($teller, 0, $branch, $filler) ;
SqlErr("load_db -- insert into teller") ;
}
print_log("teller table loaded") ;
$ update statistics for table branch ;
SqlErr("load_db -- update stats on branch") ;
$ update statistics for table teller ;
SqlErr("load_db -- update stats on teller") ;
$ close database ;
SqlErr("load_db -- close database") ;
sqlexit() ;
if (load_procs) {
beg_time = time(0) ;
startacct = skip * A_PERB ;
for (i=0; i < load_procs; i++) {
sprintf(begnum,"%d",startacct) ;
startacct += acct_hunk ;
sprintf(endnum,"%d",startacct-1) ;
pid = fork() ;
if (pid == -1) {
perror("on fork of loadact process") ;
exit(1) ;
}
if (pid == 0)
if (logfile)
execl("load_act","load_act",begnum,endnum,"1",log_fname,0) ;
else
execl("load_act","load_act",begnum,endnum,"0"," ",0) ;

C-3 TPC Benchmark B Full Disclosure

Appendix C Code to Populate Database

}
while (i--) {
pid = wait(&wait_status) ;
if (pid == -1) {
perror("on return from loadact") ;
exit(1) ;
}
end_time = time(0) ;
totsecs = end_time - beg_time ;
sprintf(rpt_str,"\nprocess %d completed in ",pid) ;
report_time(rpt_str,totsecs) ;
if (i > 0)
sprintf(rpt_str,"%s; %d procs still
working",rpt_str,i) ;
print_log(rpt_str) ;
}
sprintf(rpt_str,"\nAll

processes

finished

%s",ctime(&end_time)) ;
print_log(rpt_str) ;
sprintf(rpt_str,"loaded %d account records in ",load_accts) ;
report_time(rpt_str,totsecs) ;
sprintf(rpt_str,"%s = %d rows/sec\n",rpt_str,load_accts/totsecs)
;
print_log(rpt_str) ;
$ database tpc ;
SqlErr("Open Database") ;
$ update statistics for table account ;
SqlErr("Update Statistics on account") ;
$ close database ;
SqlErr("Close Database") ;
}
if (logfile)

fclose(flog) ;

exit(0) ;
}
report_time(s,secs)
char s[] ;
long secs ;
{
int hrs, mins, slen ;
hrs = secs / 3600 ;
secs = secs % 3600 ;
mins = secs / 60 ;
secs = secs % 60 ;
slen = strlen(s) ;
sprintf(&s[slen],"%2d:%02d:%02d",hrs,mins,secs) ;
}

TPC Benchmark B Full Disclosure

C-4

Appendix C Code to Populate Database

print_log(s)
char *s ;
{
if (logfile) {
fprintf(flog,"%s\n",s) ;
fflush(flog) ;
}
else {
printf("%s\n",s) ;
fflush(stdout) ;
}
}

C-5 TPC Benchmark B Full Disclosure

Appendix D
Database Contents Samples
This appendix contains the database contents samples for the TPC Benchmark B
run on the DECsystem 5500.
D.1 Branch Table
Following is a sample of the Branch table contents:
number
balance
fillstr

5
-6004309
1234567890123456789012345678901234567890123456789012345678
9012345678901234567890123456789012

number
balance
fillstr

6
356395575
1234567890123456789012345678901234567890123456789012345678
9012345678901234567890123456789012

number
balance
fillstr

7
-10939855
1234567890123456789012345678901234567890123456789012345678
9012345678901234567890123456789012

D.2 Teller Table
Following is a sample of the Teller table contents:
number
balance
branch
fillstr

10
-25340103
1
1234567890123456789012345678901234567890123456789012345678
9012345678901234567890123456789

number
balance
branch
fillstr

11
43345491
1
1234567890123456789012345678901234567890123456789012345678
9012345678901234567890123456789

number
balance
branch
fillstr

12
-18705273
1
1234567890123456789012345678901234567890123456789012345678
9012345678901234567890123456789

TPC Benchmark B Full Disclosure D-1

Appendix D Database Contents Samples

D.3 History Table
Following is a sample of the History table contents:
account
teller
branch
delta
tstamp
fillstr

3812872
380
38
24901
1991-11-08 13:02:11
the rest is history

account
teller
branch
delta
tstamp
fillstr

299394
85
8
36671
1991-11-08 13:02:11
the rest is history

account
teller
branch
delta
tstamp
fillstr

3032051
300
30
-833240
1991-11-08 13:02:11
the rest is history

D.4 Account Table
Following is a sample of the Account table contents:
number
balance
branch
fillstr

1244
381120
0
1234567890123456789012345678901234567890123456789012345678
90123456789012345678901234567

number
balance
branch
fillstr

1245
818867
0
1234567890123456789012345678901234567890123456789012345678
90123456789012345678901234567

number
balance
branch
fillstr

1246
-248031
0
1234567890123456789012345678901234567890123456789012345678
90123456789012345678901234567

D-2 TPC Benchmark B Full Disclosure

Appendix E
Device Configurations
This appendix contains a description of the physical disk configurations tested for the
DECsystem 5500 configuration.
/dev/rra0a
Current partition table:
partition
bottom
top
a
0
32767
b
32768
163839
c
0
744399
d
0
163839
e
0
471039
f
471040
744399
g
163840
744399
h
0
0

size
32768
131072
744400
163840
471040
273360
580560
0

overlap
c,d,e,h
c,d,e
a,b,d,e,f,g,h
a,b,c,e,h
a,b,c,d,g,h
c,g
c,e,f
a,c,d,e

/dev/rra1a
Current partition table:
partition
bottom
top
a
0
65535
b
65536
265143
c
0
744399
d
0
163839
e
0
471039
f
471040
744399
g
265144
744399
h
0
0

size
65536
199608
744400
163840
471040
273360
479256
0

overlap
c,d,e,h
c,d,e
a,b,d,e,f,g,h
a,b,c,e,h
a,b,c,d,g,h
c,g
c,e,f
a,c,d,e

/dev/rra2a
Current partition table:
partition
bottom
top
a
0
482255
b
482256
744399
c
0
744399
d
0
163839
e
0
471039
f
471040
744399
g
0
0
h
0
0

size
482256
262144
744400
163840
471040
273360
0
0a

overlap
c,d,e,f,g,h
c,f
a,b,d,e,f,g,h
a,c,e,g,h
a,c,d,g,h
a,b,c
a,c,d,e,h
,c,d,e,g

TPC Benchmark B Full Disclosure E-1

E Device Configurations

/dev/rra3a
Current partition table:
partition
bottom
top
a
0
32767
b
32768
163839
c
0
744399
d
0
163839
e
0
471039
f
471040
744399
g
163840
744399
h
0
0

size
32768
131072
744400
163840
471040
273360
580560
0

overlap
c,d,e,h
c,d,e
a,b,d,e,f,g,h
a,b,c,e,h
a,b,c,d,g,h
c,g
c,e,f
a,c,d,e

/dev/rra4a
Current partition table:
partition
bottom
top
a
0
32767
b
32768
163839
c
0
744399
d
0
163839
e
0
471039
f
471040
744399
g
163840
744399
h
0
0

size
32768
131072
744400
163840
471040
273360
580560
0

overlap
c,d,e,h
c,d,e
a,b,d,e,f,g,h
a,b,c,e,h
a,b,c,d,g,h
c,g
c,e,f
a,c,d,e

/dev/rra5a
Current partition table:
partition
bottom
top
a
0
32767
b
32768
163839
c
0
744399
d
0
163839
e
0
471039
f
471040
744399
g
163840
744399
h
0
0

size
32768
131072
744400
163840
471040
273360
580560
0

overlap
c,d,e,h
c,d,e
a,b,d,e,f,g,h
a,b,c,e,h
a,b,c,d,g,h
c,g
c,e,f
a,c,d,e

/dev/rra6a
Current partition table:
partition
bottom
top
a
0
32767
b
32768
163839
c
0
744399
d
0
163839
e
0
471039
f
471040
744399
g
163840
744399
h
0
0

size
32768
131072
744400
163840
471040
273360
580560
0

overlap
c,d,e,h
c,d,e
a,b,d,e,f,g,h
a,b,c,e,h
a,b,c,d,g,h
c,g
c,e,f
a,c,d,e

E-2 TPC Benchmark B Full Disclosure

E Device Configurations

/dev/rrz0a
No partition table found in superblock... using default table from device driver.
Current partition table:
partition
bottom
top
size
overlap
a
0
32767
32768
c
b
32768
217087
184320
c
c
0
1954049
1954050
a,b,d,e,f,g,h
d
831488
1130495
299008
c,h
e
1130496
1429503
299008
c,h
f
1429504
1954049
524546
c,h
g
217088
831487
614400
c
h
831488
1954049
1122562
c,d,e,f
/dev/rrz1a
Current partition table:
partition
bottom
top
a
0
32767
b
32768
217087
c
0
1954049
d
831488
1130495
e
1130496
1429503
f
1429504
1954049
g
217088
831487
h
831488
1954049

size
32768
184320
1954050
299008
299008
524546
614400
1122562

overlap
c
c
a,b,d,e,f,g,h
c,h
c,h
c,h
c
c,d,e,f

/dev/rrz2a
Current partition table:
partition
bottom
top
a
0
32767
b
32768
163839
c
0
1299173
d
163840
456369
e
456370
748899
f
748900
1299173
g
163840
731505
h
731506
1299173

size
32768
131072
1299174
292530
292530
550274
567666
567668

overlap
c
c
a,b,d,e,f,g,h
c,g
c,g,h
c,h
c,d,e
c,e,f

/dev/rrz3a
No partition table found in superblock... using default table from device driver.
Current partition table:
partition bottom
top
size
overlap
a
0
32767
32768
c
b
32768
163839
131072
c
c
0
1299173
1299174
a,b,d,e,f,g,h
d
163840
456369
292530
c,g
e
456370
748899
292530
c,g,h
f
748900
1299173
550274
c,h
g
163840
731505
567666
c,d,e
h
731506
1299173
567668
c,e,f

TPC Benchmark B Full Disclosure

E-3

E Device Configurations

/dev/rrz4a
Current partition table:
partition
bottom
top
a
0
32767
b
32768
163839
c
0
1299173
d
163840
456369
e
456370
748899
f
748900
1299173
g
163840
731505
h
731506
1299173

size
32768
131072
1299174
292530
292530
550274
567666
567668

E-4 TPC Benchmark B Full Disclosure

overlap
c
c
a,b,d,e,f,g,h
c,g
c,g,h
c,h
c,d,e
c,e,f

Appendix F
System Parameter Settings
This appendix contains the operating system parameters and database options in the
TPC Benchmark B test system.

F.1 System Parameters
ULTRIX version 4.2 system parameters were configured as shown below. In all instances default values were used except for
•

MAXUSERS was set to 128

•

MAXUPRC was set to 128

•

SMMAX was set to 1024

•

SMSEG was set to 128

Additionally, two semaphore constant values were changed in the ULTRIX IPC
Semaphore Facility sem.h (/usr/sys/h/sem.h). The value SEMMNI, the number of
semaphore identifiers, was set to 40, and the SEMMNS, the number of semaphores
in the system, was set to 120. A copy of sem.h appears in this appendix.
The following operating system parameters were used for the test system.
ident
machine
cpu
maxusers
processors
maxuprc
physmem
timezone
smmax
smseg

"DIMES"
mips
"DS5500"
128
1
128
32
5 dst 1
1024
128

options
options
options
options
options
options
options
options
options

LAT
QUOTA
INET
EMULFLT
NFS
RPC
DLI
NETMAN
UFS

TPC Benchmark B Full Disclosure F-1

Appendix F System Parameter Settings

F-2options

DECNET

makeoptions

ENDIAN="-EL"

config

vmunix

root on ra0a swap on ra0b dumps on ra0b

adapter
adapter
adapter
controller
disk
controller
disk
controller
disk
controller
disk
controller
disk
controller
disk
controller
disk
controller
disk
disk
disk
disk
disk
tape
device

uba0
msi0
ibus0
dssc0
ra0
dssc1
ra1
dssc2
ra2
dssc3
ra3
dssc4
ra4
dssc5
ra5
dssc6
ra6
asc0
rz0
rz1
rz2
rz3
rz4
tz6
ne0

at nexus?
at nexus?
at nexus?
at msi0
at dssc0
at msi0
at dssc1
at msi0
at dssc2
at msi0
at dssc3
at msi0
at dssc4
at msi0
at dssc5
at msi0
at dssc6
at ibus?
at asc0
at asc0
at asc0
at asc0
at asc0
at asc0
at ibus?

scs_sysid

pseudo-device
pseudo-device
pseudo-device
pseudo-device
pseudo-device
pseudo-device
pseudo-device
pseudo-device
pseudo-device
pseudo-device
pseudo-device
pseudo-device
pseudo-device

pty
loop
inet
ether
lat
lta
rpc
nfs
dli
netman
ufs
decnet
presto

F-2 TPC Benchmark A Full Disclosure

msinode 0
drive 0
msinode 1
drive 1
msinode 2
drive 2
msinode 3
drive 3
msinode4
drive 4
msinode 5
drive 5
msinode 6
drive 6
vector ascintr
drive 0
drive 1
drive 2
drive 3
drive 4
drive 6
vector neintr

Appendix F

System Parameter Settings

F.2 IPC Semaphore Facility
/* @(#)sem.h

4.1 (ULTRIX)

7/2/90 */

/************************************************************************
*
*
Copyright (c) 1986, 1988 by
*
Digital Equipment Corporation, Maynard, MA
*
All rights reserved.
*
* This software is furnished under a license and may be used and
* copied only in accordance with the terms of such license and
* with the inclusion of the above copyright notice. This
* software or any other copies thereof may not be provided or
* otherwise made available to any other person. No title to and
* ownership of the software is hereby transferred.
*
* This software is derived from software received from the
* University of California, Berkeley, and from Bell
* Laboratories. Use, duplication, or disclosure is subject to
* restrictions under license agreements with University of
* California and with AT&T.
*
* The information in this software is subject to change without
* notice and should not be construed as a commitment by Digital
* Equipment Corporation.
*
* Digital assumes no responsibility for the use or reliability
* of its software on equipment which is not supplied by Digital.
*
************************************************************************/
/* * * Modification history:
*
* 19 Mar 90 -- burns
*
Added ifdef kernel around SMP lock imbedded in
*
a user visable data structure (msqid_ds).
*
* 13 Dec 89 -- scott
*
xpg compliance changes
*
* 16 Aug 88 -- miche
*
Add support for SMP
*
* 02 Apr 86 -- depp
*
Moved sizing constants from /sys/h/param.h to here.
*
* 01 Mar 85 -- depp *
New file derived from System V IPC
*
*/

*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*

TPC Benchmark B Full Disclosure

F-3

Appendix F System Parameter Settings

/*
**
*/

IPC Semaphore Facility.

#ifndef KERNEL
#include <sys/smp_lock.h>
extern int semctl();
extern int semget();
extern int semop();
#endif /* KERNEL */
#if !defined(_POSIX_SOURCE)
/*
**
Implementation Constants.
*/
#define
#define

PSEMN
PSEMZ

(PZERO + 3)
(PZERO + 2)

/*
**
*/

Permission Definitions.

#define
#define

SEM_A
SEM_R

0200
0400

/* sleep priority waiting for greater value */
/* sleep priority waiting for zero */

/* alter permission */
/* read permission */

#endif /* !defined(_POSIX_SOURCE) */
/*
**
Semaphore Operation Flags.
*/
#define

SEM_UNDO010000

/*
**
*/

Semctl Command Definitions.

#define
#define
#define
#define
#define
#define
#define

GETNCNT
GETPID
GETVAL
GETALL
GETZCNT
SETVAL
SETALL

/*
**
*/

Structure Definitions.

/*
**
**
**
*/

There is one semaphore id data structure for each set of semaphores
in the system. The ipc_perm structure must be first and
the lock must be last.

3
4
5
6
7
8
9

/* set up adjust on exit entry */

/* get semncnt */
/* get sempid */
/* get semval */
/* get all semval’s */
/* get semzcnt */
/* set semval */
/* set all semval’s */

F-4 TPC Benchmark A Full Disclosure

Appendix F

System Parameter Settings

struct semid_ds {
struct ipc_perm
sem_perm; /* operation permission struct */
struct sem *sem_base; /* ptr to first semaphore in set */
unsigned short
sem_nsems;/* # of semaphores in set */
time_t
sem_otime; /* last semop time */
time_t
sem_ctime; /* last change time */
#ifdef KERNEL
struct __lock_t sem_lk; /* SMP lock for the semaphore queue */
#endif /* KERNEL */ };
/*
**
*/

There is one semaphore structure for each semaphore in the system.

struct sem {
unsigned short semval; /* semaphore text map address */
pid_t sempid;
/* pid of last operation */
unsigned short semncnt; /* # awaiting semval > cval */
unsigned short semzcnt; /* # awaiting semval = 0 */
unsigned short semnwakup;/* wake up those waiting on semncnt */
};
#if !defined(_POSIX_SOURCE)
/*
**
*/

There is one undo structure per process in the system.

struct sem_undo {
struct sem_undo
short
struct undo {
short
short
int
}
un_ent[1];
};
/*
** semaphore information structure
*/
struct
seminfo
{
int
semmap,
semmni,
semmns,
semmnu,
semmsl,
semopm,
semume,
semusz,
semvmx,
semaem;
};

*un_np;
un_cnt;

/* ptr to next active undo structure */
/* # of active entries */

un_aoe;
/* adjust on exit values */
un_num;
/* semaphore # */
un_id;
/* semid */
/* undo entries (one minimum) */

/* # of entries in semaphore map */
/* # of semaphore identifiers */
/* # of semaphores in system */
/* # of undo structures in system */
/* max # of semaphores per id */
/* max # of operations per semop call */
/* max # of undo entries per process */
/* size in bytes of undo structure */
/* semaphore maximum value */
/* adjust on exit max value */

TPC Benchmark B Full Disclosure

F-5

Appendix F System Parameter Settings

/*
**
*/

User semaphore template for semop system calls.

struct sembuf {
unsigned short sem_num;
short
sem_op;
short
sem_flg;
};
/*
* Sizing constants
*/
#define SEMMAP 10
#define SEMMNI 40
#define SEMMNS 120
#define SEMMNU 30
#define SEMMSL 25
#define SEMOPM 10
#define SEMUME 10
#define SEMVMX 32767
#define SEMAEM 16384
#endif /* !defined(_POSIX_SOURCE) */

F-6 TPC Benchmark A Full Disclosure

/* semaphore # */
/* semaphore operation */
/* operation flags */

Appendix G
Independent Auditor’s Report

TPC Benchmark B Full Disclosure G-1