Pegasus and the Pulsar Search: From Metadata to Execution on the Grid

摘要

This paper describes the Pegasus workflow mapping and planning system that can map complex workflows onto the Grid. In particular, Pegasus can be configured to generate an executable workflow based on application-specific attributes. In that configuration, Pegasus uses and AI-based planner to perform the mapping from high-level metadata descriptions to a workflow that can be executed on the Grid. This configuration of Pegasus was used in the context of the Laser Interferometer Gravitational Wave Observatory (LIGO) pulsar search. We conducted a successful demonstration of the system at SC 2002 during which time we ran approximately 200 pulsar searches.