Home // ICN 2011, The Tenth International Conference on Networks // View article

Modeling and Evaluation of SWAP Scheduling Policy Under Varying Job Size Distributions

Authors:
Idris A. Rai
Michael Okopa

Keywords: Size-based scheduling; conditional mean response time; temporal dependence

Abstract:
Size-based scheduling policies have been shown to be effective resource allocation policies in computing and networked environments. One of the recently proposed size-based scheduling policy is called SWAP. It is a non-preemptive, threshold based policy that was proposed to approximate the Shortest Job First (SJF) policy by introducing service differentiation between short and large jobs such that short jobs are given service priority over the large jobs. Original study of the SWAP scheduling policy was based on only simulations, which are known to have a number of restrictions. In this paper, we derive SWAP models and evaluate the scheduling policy using workloads that have varying distributions. In contrast to simulations, the models enable fast analysis of the scheduling policy under a wide range of input parameters. Numerical results obtained from the derived models show that SWAP approximates SJF better for heavy-tailed workloads than for exponentially distributed workloads. We also show that SWAP performs significantly better than First Come First Serve (FCFS) and Processor Sharing (PS) policies regardless of the distribution of the workload.

Pages: 115 to 120

Copyright: Copyright (c) IARIA, 2011

Publication date: January 23, 2011

Published in: conference

ISSN: 2308-4413

ISBN: 978-1-61208-113-7

Location: St. Maarten, The Netherlands Antilles

Dates: from January 23, 2011 to January 28, 2011