Pegasus on Virtual Grid: A Case Study of Workflow Planning over Captive Resources

As scientific applications in a variety of disciplines are being actively studied and computing environments become more complicated and dynamic, a challenge is how to leverage existing cyber-infrastructures and achieve robust and efficient computing. This paper presents a case study on planning and executing application workflows over provisioned resources. This work integrates the Pegasus workflow framework with the Virtual Grid resource provisioning system. Through this preliminary study, we have identified two key issues that need to be addressed: 1) resource capacity estimate which synthesizes efficient resource descriptions for Virtual Grid from application workflows and 2) execution site information publication which generates resource configurations for Pegasus through the devirtualization of provisioned resources. In addition, we discuss the challenges and research opportunities that we need to explore in order to exploit the advanced features of the systems.