Time/Cost Optimization in Grid Scheduling: Analytical Results and Practical Implications

An important aspect of Grid scheduling is efficient allocation of computation tasks to Grid resources. Recent trends in Grid computing involve Quality of Service provision which ensures that an application is performed within a guaranteed time at a pre-specified cost. We develop efficient algorithms for finding a trade-off between computation time and resource usage cost which gives resource providers and users a range of Quality-of-Service options to select from. Theoretical outcomes obtained for simplified models are generalized for a more complex model of scheduling workflows. Computational experiments demonstrate that the rule that prioritizes the processors in accordance with the cost-over-speed ratio outperforms the rule which considers the cost and the speed parameters separately.