A Method for Performance Analysis of Earliest-Deadline-First Scheduling Policy
The Journal of Supercomputing
Hi-index | 0.00 |
This paper introduces an analytical method forapproximating the fraction of jobs that miss theirdeadlines in a real-time system when earliest-deadline-firstscheduling policy (EDF) is used. In the system, jobseither all have deadlines until the beginning of service ordeadlines until the end of service. In the former case,EDF is known to be optimal and, in the latter case, it willbe optimal if preemption is allowed. In both cases, thesystem is modeled by an M/M/1/EDF+M queue, i.e., asingle server queue with Poisson arrival, and servicetimes and customer impatience, which are exponentiallydistributed. The optimality property of EDF is used forthe estimation of a key parameter, 驴{n}, which is the lossrate when there are n customers in the system. Theestimation is possible by finding an upper bound and alower bound for 驴{n} and linearly combining these twobounds. The resulting Markov chains are then easy tosolve numerically. Comparing numerical and simulationresults, we find that the existing errors are relatively small.