dtj v01-05 sep1987

This document is Number 5 of the Digital Technical Journal, published in September 1987, and is dedicated to VAXcluster Systems.

The journal issue explores how VAXcluster systems, described as closely coupled configurations of VAX CPUs and storage devices, operate as a single distributed system, emphasizing their design for high availability and extensibility.

Key areas covered in the issue include:

1. The VAXcluster Concept Overview: An architectural overview detailing how the combination of the VAX/VMS operating system (with its distributed lock manager), high-speed message-oriented Computer Interconnect (CI bus), and intelligent CI port hardware enables transparent data sharing and high performance across multiple nodes.
2. System Communication Architecture (SCA): The underlying framework that allows VAXcluster nodes to cooperate, facilitating data sharing and binding applications across different VAX CPUs with low overhead.
3. VAX/VMS Distributed Lock Manager: The core synchronization mechanism enabling transparent and reliable data sharing by managing access to shared resources across the cluster, including deadlock detection and reconfiguration handling.
4. Distributed File System: The design and implementation of the file system, which leverages the distributed lock manager and an extended QIO processor to synchronize file accesses and manage caches across the cluster, ensuring data consistency.
5. Local Area VAXcluster (LAVC) Systems: The extension of the VAXcluster concept to Ethernet, allowing smaller systems like MicroVAX and VAXstation to benefit from cluster functionality through software emulation of CI bus functions for remote disk access and network booting.
6. Performance and Availability Modeling: Papers analyze the performance aspects of VAXcluster systems, including:
   • Availability Modeling: Demonstrating how hardware redundancy (processors, storage controllers, disks) significantly improves system availability, reducing downtime.
   • System-Level Performance: Evaluating the performance of VAX 8974 and 8978 systems in scientific and transaction processing environments, comparing their throughput to a single VAX 8700 CPU and discussing factors like communication and locking overhead.
   • CI Bus Arbitration Performance: Studying the arbitration algorithm of the CI bus using a generalized semi-Markov process model to understand its characteristics and suggest ways to improve bus utilization and reduce response time.

Overall, the journal highlights the hardware and software innovations that enabled VAXcluster systems to deliver a highly integrated, scalable, and resilient distributed computing environment.