AN EFFECTIVE PROCESSOR ALLOCATION STRATEGY FOR MULTIPROGRAMMED SHARED-MEMORY MULTIPROCESSORS

Existing techniques for sharing the processing resources in multiprogr ammed shared-memory multiprocessors, such as time-sharing, space-shari ng, and gang-scheduling, typically sacrifice the performance of indivi dual parallel applications to improve overall system utilization. We p resent a new processor allocation technique called Loop-Level Process Control (LLPC) that dynamically adjusts the number of processors an ap plication is allowed to use for the execution of each parallel section of code, based on the current system load. This approach exploits the maximum parallelism possible for each application without overloading the system. We implement our scheme on a Silicon Graphics Challenge m ultiprocessor system and evaluate its performance using applications f rom the Perfect Club benchmark suite and synthetic benchmarks. Our app roach shows significant improvements over traditional time-sharing and gang-scheduling. It has performance comparable to, or slightly better than, static space-sharing, but our strategy is more robust since, un like static space-sharing, it does not require a priori knowledge of t he applications' parallelism characteristics.