Goodwill based scheduling algorithm for economy grid

Abstract

The conventional deadline and budget constrained scheduling heuristics for economics-based computational grids do not consider goodwill of resources, that could lead to an increase in application cost and execution time. This paper proposes a new scheduling algorithm that not only considers resource cost and time but also considers goodwill when resources are selected. Resources goodwill is calculated based on their past performances. Goodwill is broader and more representative characteristic than reliability of a resource to use as criterion for selecting the resource for execution. Goodwill criterion takes into account “on time performance” of a resource calculated as real time basis where reliability criterion works on a predefined static term expressed by the system designer. The Gridsim toolkit, standard workload model and resource configuration was used to simulate the grid environment and application. The experimental results show that the average overall time and cost of parallel tasks are lesser than existing approaches.