A Quality-Driven Algorithm for Resource Scheduling Based on Market Model on Grid

Several challenges about computational grid exist in integrating, coordinating and managing of resources and scheduling of applications, due to distributed resources at various levels. In this paper, we employ computational economy as a way to manage and allocate these distributed resources. It can help to rapidly and dynamically select resources meeting users' functional requirements and satisfying their Quality of Service (QoS), from a large number of resources. An economic scheduling system based on market model is presented. On the basis of studying QoS criteria of resource services, we formulate selection as Multiple Attribute Decision Making (MADM) problem. Then, we introduce MQoS model to evaluate resource services for selection, which consists of the MQoS vector, a decision matrix and a quality- driven scheduling algorithm. The algorithm is classified as lexicographic and Euclidean Distance algorithm with respect to user's preference and object impact. It optimizes system performance through adjusting resource selection automatically according to previous job execution and the current system state information such as load. The results of scheduling experiments showed the superiority of quality-driven algorithm in scheduling on global grids.