VDS - The GriPhyN Virtual Data System

(formerly, "Chimera", and including "Pegasus")

VDS, The Virtual Data System, provides a set of tools for expressing, executing, and tracking the results of workflows.

Workflows consist of graphs of application (or soon, service) invocations, and can be expressed in a location-independent, high-level "Virtual Data Language" - VDL - which free's the workflow from specifiying details of the location of files and programs in a distributed environment.

(Information on the Pegasus Planner and the latest code release of Pegasus 2.0 successor to Pegasus 1.4.8 (Included in VDS 1.4.8) is available at http://pegasus.isi.edu).

(Information on the Swift workflow system and the SwiftScript language for loosely coupled parallel computation, the sucessor to VDL, is here - www.ci.uchicago.edu/swift).

VDL workflows can be executed in a variety of environments ranging from the desktop to Grids such as the Open Science Grid and Teragrid.

VDL defintions are stored in a Virtual Data Catalog - VDC - that provides for the tracking of the provenance of all files derived by the workflow.

Both VDL and the workflows that it generates can be expressed in XML documents. The workflow is expressed as a "DAG in XML" or "DAX" document, and is used by planner components in the VDS to generate executable forms of the workflow.

VDS has a primary planner named Pegasus, and an experimental research planner that is being used to prototype new planning features and executable code generation approaches. The planners have the following characteristics:

Pegasus - production planner that can generate a "fully planned" executable workflow, and various styles of "incrementally planned" executable workflows. It produces Condor DAGman DAGs as output. Pegasus is available as a standalone distribution now starting with version 2.0 here that can be interfaced to a variety of workflow composition tools including Wings, VDL , DAX Api etc. More information about Pegasus, the latest code, and ongoing research can be found at http://pegasus.isi.edu

Euryale - an experimental incremental planner that generates a Condor DAGman DAG in which sites are selected for a job when the job is determined by DAGman to be "ready to run". This planner uses a template-driven approach to make the Grid-specific aspects of each job very flexible.

Pegasus the production planner in VDS 1.4.8 is now available as PEGASUS 2.0.0 in a standalone release here. For any planner related questionS in VDS OR Pegasus contact pegasus-discuss at isi dot edu or pegasus at isi dot edu

The last stable release of VDS was 1.4.8 and available in the archives of the Download page.