VAXcluster Systems Quorum

This document is a technical journal focusing on VAXcluster systems management, comprising three main articles and a customer questionnaire.

The first article, "About the HSC Cache," explains how the VAXcluster High-Speed Cache (HSC) works. It details that the 32MB ECC memory cache reduces disk access response times for I/O streams characterized by frequent, small, repetitive reads, thus boosting throughput in timeshare environments. It differentiates HSC cache from Solid-State Disks (ESEs), noting they are complementary. The article lists types of storage not cached (e.g., tape, Phase I shadowed disks, ESEs, disks with data check) and provides guidance on how to assess if an HSC cache will benefit a system using the Cache Needs Analysis Tool (CNAT). It also covers hardware requirements, how the PDP-11 processor accesses cached blocks, and commands for enabling, configuring, and verifying cache settings (e.g., SETSHO, VTDPY, DSCACHE, DSTAT). Various write allocation policies (ALLOCATE, INVALIDATE, UPDATE) are described for managing how write requests interact with the cache.

The second article, "VAXcluster State Change Timing," focuses on the recovery mechanisms in VAXcluster systems during "transitions" or state changes caused by connectivity disturbances. It outlines seven phases: failure detection, repair attempt, last gasp processing, new member detection, reconfiguration, system recovery, and application recovery. Key parameters like TIMVCFAIL (for detection time) and RECNXINTERVAL (for repair attempts) are explained, along with their impact on transition duration and system stability. The article details lock database rebuild types (Merge, Partial, Directory) and quorum disk validation as critical system recovery tasks, noting their effect on user activity. It concludes with a summary table of typical transition times for different interconnects and suggestions for controlling these timings, such as avoiding node halts and deferring disk rebuilds.

The third article, "Building Large VAXcluster Configurations," addresses the complexities and solutions for setting up VAXcluster systems with a large number of nodes. It highlights benefits like shared resource access and easier system management. The article discusses technological advancements that facilitate larger configurations, including faster CPUs, improved network adapters (FDDI, DSSI), and VAXcluster software enhancements. It delves into critical configuration factors such as optimizing boot times (e.g., offloading system disk, avoiding disk rebuilds with MOUNT/NOREBUILD and ACP_REBLDSYSD, utilizing caching with HSC/ESEs/DECram), ensuring sufficient network bandwidth (e.g., multiple segments, FDDI backbone, DECbootsync), and managing system parameters (SCSCONNCNT, SCSRESPCNT, SCSBUFFCNT). It also covers strategies for protecting the network, managing hot system files, ensuring MOP/Disk server availability, and dealing with cluster alias limits.

Finally, a "VAXcluster Customer Configuration Database Questionnaire" is included, designed to collect detailed customer configuration data to identify trends, forecast product needs, and improve customer service. It requests information on system hardware, software, applications, and customer contacts, with options for data security.