Dynamic Task Distribution in the Grid for BLAST

Abstract

BLAST programs are being widely used as tools for searching protein and DNA databases for sequence similarities. Even though many parallelization approaches have been developed in order to decrease the search time, individual databases are growing at a faster pace making searches longer than ever. Fortunately, Grid middleware is finding its acceptance across many campus Grids which are starting to utilize more and more distributed and smaller resources which would normally sit idle. Dynamic BLAST is an open-source implementation relying on Grid middleware technologies to distribute multiple queries across many, possibly geographically distributed, resources and thus perform searches resulting in shorter turnaround time. All of the intricacies of resource acquisition, data distribution, and job submission are handled by Dynamic BLAST, without any user intervention. We present the architectural and implementation details along with the experimental results to illustrate the scalability and usability of the system in a campus grid.