AUCTION BASED RESOURCE ALLOCATION IN COMPUTATIONAL GRIDS

Abstract

In this Dissertation work, several auction based resource allocation policies proposed for Grid users, which assist them in choosing the resource according to user preference. While choosing the resource these policies try to optimize parameters like average turnaround time, average budget per job, and number of jobs finished within deadline according to user preference. Initially proposed simple Time Optimized and Cost Optimized policies which improve only one parameter i.e. either average turnaround time or average cost per job. Compared these policies with a Random policy which selects resources randomly. Later improved these algorithms and proposed New Time Optimized and New Cost Optimized policies which consider the success rate also. Finally proposed the Adaptive policy that tries to optimize above parameters by considering the success rate of tasks submitted by user. Adaptive Policy initially uses time optimization policy once user specified success rate reaches it shifts to cost optimization policy. GridSim simulation framework used to evaluate resource allocation policies and for implementation of auction mechanisms.