Data Communication Model of Grid Workflow

Many grid workflow applications, especially data intensive applications, need to transmit large amount of data between grid services. Centralized data transmission adopted in most of classical workflow systems, i.e. the workflow engines delay all data transmitting across the services, would results in a bottleneck of communication in these grid workflows. This paper proposes a decentralized data communication model that enables peer-to-peer communication for large amount of data between grid services by means of notification mechanism of grid service and GridFTP. To enable inter-workflow data communication in the case of composite grid workflow that is composed of multiple grid workflows, data proxy services are constructed to take charge of data communication tasks of workflows. After analyzing a number of data communication patterns in the composite grid workflow, an example is given to justify the use of data proxy services