A model for resource management in computational grid for real-time jobs using game theory

Abstract

In a computational grid system, resources lie under the purview of different administrative domains with varying policies for their usages. Resource management, thus, is an important and core issue for making such a grid system viable. The resource provider may set different resource usage policies depending on the current system load and resource availability at that point of time, whereas the resource consumer may also insist upon the quality of service for its job. For real-time jobs, resource consumer may seek guaranteed response within strict time frames. The proposed work first identifies the grid clusters based on different users' parameters viz. cost, execution time and reliability before the resource consumer and provider start playing a non-cooperative game in which players act independently to choose their actions in a sealed bid auction. Thus, based on the outcome of the game, the highest bidder in the game is chosen and accordingly resource allocation is done.