Ensemble Broker Service Oriented Architecture for LEAD

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Wednesday, 1 February 2006: 11:00 AM
Ensemble Broker Service Oriented Architecture for LEAD
A412 (Georgia World Congress Center)
Jay Alameda, National Center for Supercomputing Applications, Champaign, IL; and S. Hampton, B. F. Jewett, A. Rossi, and R. B. Wilhelmson

Presentation PDF (741.5 kB)

We will be demonstrating a set of services designed to broker large number of computational jobs, as described in requirements for the Linked Environments for Atmospheric Discovery (LEAD) ITR project. This broker system, in particular, will have the ability to broker on-demand jobs of various urgency levels, through an on-demand service being developed in collaboration with Cluster Resources, Inc. The broker is designed to support the launching of a multiplicity of ensembles of workflows. An ensemble, for our purpose, is a collection of workflows; each workflow is comprised of nodes of a variety of flavors. These flavors could include remote invocations of our compute-node-local workflow engine, the Open Grid computing environments Runtime Engine (OGRE), a service invocation, or even a plain script. Within a workflow description, we can describe properties and constraints for example, we may want to place groups of nodes on the same computational resource, or a group of nodes (or workflows, or an ensemble) may need to be run in some on-demand fashion. We will be demonstrating early capabilities of the broker system by executing a subset of the full LEAD capabilities, as motivated by ensemble requirements from the atmospheric science community, on production NCSA and TeraGrid resources.