The performance of local and global scheduling strategies in multiprogrammed parallel systems

Abstract

Based on analytic models this research concentrates on basic issues that are important in the context of processor scheduling in multiprogrammed parallel systems. Local scheduling policies for fork and join jobs are analyzed. Demonstration of the appropriate multiprogramming level, and the computation of the number of processors to be allocated to a job are investigated in the context of global scheduling. The global scheduling approach is static in the sense that a fixed set of processors is allocated to a job and dynamic processor switching from one application to another is not allowed. Important insights into system behavior that are useful in the context of operating systems for parallel systems are obtained as a result of this analysis. A number of basic issues that concern the relationship between job characteristics, system performance, and effective scheduling are discussed.<>