Analysis of scheduling algorithms for proportionate fairness

Abstract

We consider a multiprocessor operating system in which each current job is guaranteed a given proportion over time of the total processor capacity. A scheduling algorithm allocates units of processor time to appropriate jobs at each time step. We measure the goodness of such a scheduler by the maximum amount by which the cumulative processor time for any job ever falls below the “fair” proportion guaranteed in the long term. In particular we focus our attention on very simple schedulers which impose minimal computational overheads on the operating system. For several such schedulers we obtain upper and lower bounds on their deviations from fairness. The scheduling quality which is achieved depends quite considerably on the relative processor proportions required by each job. We will outline the proofs of some of the upper and lower bounds, both for the unrestricted problem and for restricted versions where constraints are imposed on the processor proportions. Many problems remain to be investigated and we will give the results of some exploratory simulations. This is joint research with Micah Adler, Petra Berenbrink, Tom Friedetzky, Leslie Ann Goldberg and Paul Goldberg.