A new service reservation approach for workflow management

Many of large-scale scientific applications executed on present-day Grids, such as bioinformatics and computational chemistry, are expressed as complex workflows. To enhance Grid computing paradigm, Web Service has emerged as the de-facto communication mechanism in Grid environments. However, in heterogeneous and dynamic Grid environments, Web Services of the same type and similar functionalities are usually provided by different administrative domains and with different capabilities. This problem makes it difficult to combine suitable and effective Web Services in workflow management. In this paper, we discuss issues such as automatically generating job workflow for Grid and scheduling optimization based on scientists' requirements in context of workflow management system. First, we present a comprehensive workflow management framework in order to automatically map scientific applications to workflow processes. Then we propose a Multiple-object Candidate Algorithm based on a tree-scheduling schema to generate best-effort candidate services for each task in a workflow process. The advantage of this approach is to select Web Services which maximize user's satisfaction as well as ensure workflow optimism accordingly. We have deployed and developed this framework in the Computational Chemistry Grid. The experimental results show that our proposal is feasible and effective.