Sum-of-the-Parts Valuation-Based Scheduling of Parallel Applications over Global Grid

Grid schedulers efficiently map grid jobs to resources over the global grid network. The existing double auction-based meta-schedulers productively schedule jobs over the resources of grids. However, they are unable to address the problem of starvation in the resource-allocation process. A novel valuation method, which works along with the double auction meta-scheduler, is proposed to facilitate the adjusted utilization of resources over a grid. The two-valuation metrics are designed on the basis of Sum-of-the-Parts valuation method that aids user requirements and availability of computational resources, thus reduces the job starvation. The formal analysis is performed through Petri nets to analyze the correctness of the presented method. The comparison of the proposed valuation method and Multi-Attribute Utility Theory (MAUT)-based multiplicative valuation method is performed. The experiment results show that the proposed valuation method facilitates up to 17% more utilization of grid resources as compared to the MAUT - based multiplicative valuation method.