OpenVMS Alpha Partitioning and Galaxy Guide

This document, the "OpenVMS Alpha Partitioning and Galaxy Guide" for OpenVMS Alpha Version 7.3 (April 2001), provides comprehensive information for system managers, application programmers, and technical consultants on utilizing partitioning capabilities, OpenVMS Galaxy, and Resource Affinity Domains (RADs) on AlphaServer systems. It applies exclusively to OpenVMS Alpha systems.

Key Concepts:

1. Partitioning:

   • Hard Partitions: Physically separate computing resources with hardware-enforced barriers, preventing resource sharing. They must be on Quad Building Block (QBB) boundaries.
   • Soft Partitions (OpenVMS Galaxy): Logically separate computing resources with software-controlled access barriers, allowing dynamic reassignment of CPUs, memory, and I/O among instances. Multiple soft partitions can exist within a single hard partition.

2. OpenVMS Galaxy:

   • Allows multiple instances of the OpenVMS operating system to run on a single computer, dynamically reassigning resources like CPUs, memory, and I/O.
   • Supports different computing models: Shared-Nothing (instances fully isolated), Shared-Partial (share some resources via shared memory), and Shared-Everything (instances cluster and fully share all resources).
   • A Single-Instance Galaxy can be set up on any Alpha system (even non-Galaxy hardware) for development or evaluation, using a configuration file to define shared memory and Galaxy services.

3. Resource Affinity Domains (RADs):

   • OpenVMS Alpha systems, particularly AlphaServer GS series (where a RAD corresponds to a QBB), feature Nonuniform Memory Access (NUMA) architecture. RADs are software groupings of hardware components (CPUs, memory, I/O) with common access characteristics.
   • OpenVMS is NUMA-aware, optimizing performance by assigning processes a "home" RAD for faster local memory access, replicating read-only system code, and distributing global/reserved memory across RADs.

Features and Benefits (OpenVMS Alpha Version 7.3):

• Scalability & Adaptability: Run multiple instances (e.g., up to 8 on GS320, 2 on ES40/4100); dynamically reassign CPUs between instances without rebooting.
• Availability: Independent booting and shutdown of instances, and the ability to test new software versions in isolation.
• Performance: Leverage shared memory for high-performance communication and eliminate SMP/cluster bottlenecks.
• Clustering: Instances can be clustered within the same Galaxy (using shared memory) or with non-Galaxy systems.
• Cost of Ownership: Reduces the need for multiple physical computers, simplifying management and reducing floor space.
• Ideal for: High-availability applications like database servers, transaction processing, data warehousing, data mining, and Internet servers.

Management & Development:

• Galaxy Configuration Utility (GCU): A DECwindows Motif application that provides a graphical interface for configuring, managing, and monitoring OpenVMS Galaxy environments. It allows drag-and-drop resource reassignment and creation of active/offline configuration models.
• DCL Commands: Commands such as STOP/CPU/MIGRATE, SHOW CPU, SET CPU, SHOW MEMORY, and CONFIGURE GALAXY are available for command-line management.
• Installation Procedures: Detailed, step-by-step guides are provided for creating OpenVMS Galaxy environments on various AlphaServer models (e.g., 8400, 8200, 4100, ES40, GS80/160/320), covering hardware setup, firmware upgrades, setting console environment variables, and booting.
• Inter-Instance Communication: Features the Shared Memory Cluster Interconnect (SMCI) for high-performance communication between Galaxy instances, and a LAN Shared Memory Device Driver for Ethernet LAN emulation.
• Application Programming Interfaces (APIs): Developers can use specific APIs for locking, system event notification, and creating/managing Galaxywide shared sections (shared memory segments).
• System Dump Analyzer (SDA) Support: SDA has been enhanced to provide information specific to OpenVMS Galaxy and RADs.
• An OpenVMS Galaxy CPU Load Balancer Program is provided as an example of a privileged application that dynamically reassigns CPU resources using Galaxy services.

The guide also offers practical tips and console hints for managing OpenVMS Galaxy environments, including cautions regarding firmware updates and console commands.