UT/ORNL Bioinformatics Summit 2003 Abstract

Sequence Alignment On the Computational Grid

Asim YarKhan and Jack Dongarra

In the GrADS project, we are working on a Grid framework intended to
make it possible for users to use volatile Grid resources in an
transparent and efficient manner. As a demonstration of this
framework, the parallel sequence alignment application FastA was
adapted to use distributed, possibly replicated databases. A sequence
alignment application was chosen for our demonstration because the
size of sequence databases is growing rapidly, and it would be
preferable if the computation migrated to the data. The GrADS
framework is used to schedule and execute FastA on a Grid testbed.
The scheduling process uses an application specific performance model
to select a near-optimal set of Grid resources, given that the Grid
resources may only contain subsets of the required databases. This
work demonstrates the ability of GrADS framework to transparently
handle the details involved in executing an database application on a
computational Grid.