A model for resource management in computational grid using sequential auction and bargaining procurement

Resources in a computational grid system fall under the purview of different administrative domains of varying policies for their usages. Commercial grid offers their services (resources) on use-and-pay basis. Resource management in computational grid, offers a market place for the two prominent grid market players i.e. resource provider and resource consumer. It has been observed that, in the grid, the request for the resources may not be uniform throughout. It fluctuates from very high demand at peak time to low or negligible at off-peak time. This information may be used to fetch the resource utilization and cost benefits out of the grid. The provider would prefer to charge extra for its resources at peak time, whereas users may shift their resources usage preference to off-peak time. This work proposes a model in which the resource provider and consumer play a non-cooperative game at different time-zones and act independently to choose their actions. Some characteristic parameters such as cost, execution time and reliability have been considered to facilitate the job execution. Based on the outcome of the game, the grid cluster offering maximum reliability within the desired execution time and/or cost for the job execution. The model has been simulated for performance evaluation with quite encouraging results.