Group Management: An Approach to Processor Sharing for Cooperative Multi-task Applications

While peripheral sharing is now quite common in both commercial and experimental distributed systems, processor sharing is not. The goal of processor sharing is to allow busy users in a network to take advantage of the processing power of idle machines. Numerous experimental systems have demonstrated the usefulness of the concept, but actual use of processor sharing is still quite limited. This paper examines the question of why processor sharing has not yet become common, and presents a new approach to processor sharing, called group management. The essential idea of group management is to provide OS-level recognition of, and support for, the groups of tasks that make up processor sharing applications. With this support, writing processor sharing applications is simplified, and the utility of processor sharing is increased.