Policy-based hybrid workflow management system for QoS constrained grid application

Abstract

Grid computing is the emerging technology of sharing and clustering of heterogeneous and distributed resources. The Grid might consist of computing resources, storage resources, and various application services. As computer and network technology grows, science application is become more complex to solve the problems within expected time and workflow management systems in Grid computing are becoming more important for scientific application to solve sophisticated problem such as climate estimation, high energy physics, and bio-informatics. Efforts to solve the above problems are being made in various research centers. Although those systems can provide workflow management functionality in Grid environment, consideration of user requirements (user SLA) such as performance, reliability and interaction with user is missing. Another limitation in conventional workflow management system is lack of supporting resource management function. As each task in Grid workflow allocated to specific resource, consideration of resource management issue in workflow is necessary to guarantee various user requirements from different domain. In this thesis, we discuss how to guarantee different user requirements according to user SLA in Grid workflow management system. A hybrid workflow management system for QoS constrained Grid application is proposed to provide QoS awareness Grid workflow management system. The main contributions are an integration of existing workflow management system, a policy-based management mechanism with a CIM based policy decision scheme, integration module for workflow management and scheduling, and a prototype implementation of workflow enabled Physio-Grid application. At last, to evaluate our proposed system, we have done experiments with ECG Analysis Grid application in Physio-Grid to show the characteristics and advantages of our system compare to the existing workflow management system. The result shows that our proposed system can flexibly enhanc...