Flexible Scientific Workflows with Dynamic Embedding

Speaker: Nicholas Haasch

Time: 12:30pm-1:30pm, September 15th, 2006

Location: Nueces Conference Room

Abstract:

Scientific workflows are increasingly being used by scientists to construct and execute complex scientific analyses.  In such workflows there is typically a variety of methods to perform a single task.  However, a specific implementation for a single task must be selected before the execution of a workflow.  Selecting a specific implementation is not obvious because run time conditions can effect which implementation is best for the task.  We present an approach that allows dynamic embedding in scientific workflows.  This approach enables flexible workflows by encapsulating various implementations for a single task into an abstract component that can examine run time conditions and then execute the most appropriate implementation.  Our approach allows adaptive and flexible scientific workflows to be constructed and run within the dataflow process network that comes with built-in support for stream-based and concurrent execution.