Overview of Scientific Workflow Automation with Kepler/SPA
Speaker: Bertram Ludaescher (UC Davis) & Ilkay Altintas (San
Diego Supercomputing Center)
Title: Overview of Scientific Workflow Automation with Kepler/SPA
Abstract: The DOE SDM/SPA team and various other groups from the Kepler
community are working on an integrated problem-solving environment for
scientific workflow management and automation. A common goal of the
Kepler collaboration is to support the whole life cycle of scientific
workflows (SWFs), from modeling and design, to deployment, execution,
monitoring, and SWF exchange, reuse, and archival.
We provide an overview of (i) the characteristics and requirements of
SWFs that we encountered in a number of existing applications, (ii) the
current capabilities of Kepler in support of those requirements, and
(iii) upcoming extensions. In the SPA context, the focus is on
compute-intensive
and data-intensive workflows. Our system demonstrations highlight the
use of external job schedulers (such as NIMROD or CONDOR) to manage
compute-intensive tasks in SWFs, and the use of the SDSC Storage
Resource Broker (SRB) for data-intensive tasks, including efficient
parallel data transport, replication, and metadata-based querying of
data collections.
Last modified: Wed Jun 1 11:22:58 EDT 2005