Dynamic Cost Minimization for Multi-Class Jobs in Computational Grids

摘要

Computational grid systems usually have heterogeneous computing resources managed by different owners distributed over larger geographical areas resulting in considerable communication delays. Hence, efficient job allocation to the computing resources for improving the system performance is essential and a complex problem in these systems. In this paper, we propose a dynamic price-based job allocation scheme for multi-class jobs in computational grid systems. The objective of the proposed scheme is to provide a system-optimal solution so that the expected cost for the execution of all the jobs in the grid system is minimized. The prices charged by the computing resource owners for executing the grid users jobs are obtained based on a non-cooperative bargaining game theory framework. The performance of the proposed scheme is compared with other existing schemes using simulations with various system loads and parameters.