Analysis of Scheduling Policies under Correlated Job Sizes (CMU-CS-10-107)

Correlations in traffic patterns are an important facet of the workloads faced by real systems, andone that has far-reaching consequences on the performance and optimization of the systems involved.While there has been considerable amount of work on understanding the effect of correlations be-tween successive interarrival times, there is very little analytical work in understanding the effectof correlations between successive service requirements (job sizes). All the prior work on analyzingthe effects of correlated job sizes is limited to First-Come-First-Served scheduling. This leaves openfundamental questions such as: How do various scheduling policies interact with correlated job sizes?Can scheduling be used to mitigate the harmful effects of correlations?In this paper we take the first step towards answering these questions. We assume a simple andintuitive model for job size correlations and present the first asymptotic analysis of various commonsize-independent scheduling policies when the job size sequence exhibits high correlation. Our analy-sis reveals that the characteristics of various scheduling policies, as well as their performance relativeto each other, are markedly different under the assumption of i.i.d. job sizes versus correlated jobsizes. Further, among the class of size-independent scheduling policies, there is no single schedulingpolicy that is optimal for all degrees of correlations and thus any optimal policy must learn thecorrelations. We support the asymptotic analysis with numerical algorithms for exact performanceanalysis under an arbitrary degree of correlation, with simulations, and finally verify the lessons fromour correlation model on real world traces.