An analysis of variation in response time of CPU scheduling algorithms as a function of load: a simulation

This thesis analyzes a group of cpu scheduling algorithms on the basis of the variation in response time that results from changes in the system load. The results of this study quantify the differential degradation of performance across job categories. The job categories include short-burst interactive jobs as well as cpu intensive jobs. For each job type, measurements were made of average job turn-around time, weighted average turn-around time, and worst case response time. Additional statistics gathered include: ready-to-run queue size, cpu utilization and throughput. The three cpu scheduling algorithms compared are round-robin, shortest-job-first, and a multi-queue priority scheduler. The analysis utilizes a model encoded in ‘C’ which simulates an interactive time-sharing user community. The model allows scheduling algorithms to be measured with a controlled workload. The workload is varied by selecting the number of simulated users who are sharing the cpu