Cost Based Optimization of Job Allocation in Computational Grids

Abstract

Computational grids are distributed systems composed of heterogeneous computing resources which are distributed geographically and administratively. These highly scalable
systems are designed to meet the large computational demands of many users from scientific and business orientations. Grid computing is a powerful concept, its chief appeal being the ability to make sure all
of a resource’s computing power is used. In a grid world, the idle time of hundreds or thousands of resources could be harnessed and rented out to anyone who needed a massive infusion of processing power.
First, the architecture of a grid system is presented. The design gives a mathematical model of the grid system for efficiently allocating the grids resources. The challenges faced for optimal job allocation motivate the exploration in optimizing grid resource allocations. We have
extensively surveyed the current state of art in this area. A grid server coordinates the job allocation for the grid users and helps to select the best resources for a job among different possible resource offers with the best prices offered. Interaction between grid users and the resources require a mediator that uses different paradigm to communicate the needs of the two parties in terms of performance requirements, timing constraints, price charged etc. A game theoretic bargaining approach is studied to agree upon standard prices. We have implemented various job allocation schemes in computational grids based on the mathematical modeling of the grid system and bargaining protocol with the objective function of optimizing the cost. The performance of the schemes have been analyzed and compared. A new model for job allocation in computational grids has been proposed, for job allocation based on the clustering of resources.