THE INFLUENCE OF MULTIPROGRAMMING LIMIT ON INTERACTIVE RESPONSE TIME IN A VIRTUAL MEMORY SYSTEM

Using an analytic queueing network model with limited multiprogramming level and constrained memory, we study the variations in interactive response times due to changes in multiprogramming limit. We examine both open models (infinite user population) and closed models (finite user population). In order to measure response time meaningfully, we include the time spent gueueing for entry to the multiprogramming mix. It is found that in open systems and closed systems with a large user population, the expected time-in-system is very sensitive to multiprogramming limit. A slight deviation from the optimal value may have substantial effect. Our calculations are based upon actual lifetime functions obtained from the analysis of program reference traces.