Virtual Data System on distributed virtual machines in computational grids

This paper presents the work of building a Grid workflow system on distributed virtual machines. A Grid Virtualization Engine (GVE) is implemented to manage virtual machines as computing resources for Grid applications. The Virtual Data System (VDS) functions as a Grid workflow engine. This paper designs and implements the Virtual Data System on distributed virtual machines, which are enabled by the Grid Virtualization Engine. Various interfaces between GVE and VDS have been defined and implemented. A high energy physics application, the Compact Muon Solenoid (CMS) benchmark, is tested on a test bed equipped with the developed technologies. System discussion, test results and performance evaluation justify the paradigm of building VDS on distributed virtual machines and the prototype of our design and implementation.