SCIENCE

Pervasive Software has announced along with SGI that Pervasive's implementation of the Smith-Waterman algorithm for performing local protein sequence alignment analyzed 10 million combinations of protein sequences in 81.1 seconds. The algorithm leveraged the Pervasive DataRush platform to scale across an SGI Altix UV 1000 with 384 cores to achieve a sustained throughput of 986 billion cell updates/second. Dramatically demonstrating Pervasive DataRush scalability, the same executable used on a four-core system dynamically scaled across all 384 cores on the Intel Xeon processor 7500 series based server.1

The results provide TCUPS-scale performance, an order of magnitude greater throughput than earlier Smith-Waterman algorithm performance records. The high-performance algorithm is implemented here using the Pervasive DataRush parallel dataflow platform, which developers can rapidly exploit using its extensible library of operators that seamlessly scale on commercially available multicore hardware.

"Pervasive DataRush applications deliver robust future-proofing as customers can seamlessly scale to take advantage of ever more powerful hardware platforms, and the Smith-Waterman results provide a riveting demonstration of just how seamless and scalable that future-proofing really is," said Ray Newmark, Pervasive DataRush vice president of sales and marketing. "This demonstrates that developers in all types of organizations can leverage this platform to design and deploy highly scalable applications with no need for specialized parallelization knowledge."

"With Pervasive DataRush on Altix UV, commercial and academic researchers have unparalleled access to world-class performance for design-time and runtime productivity," said Bill Mannel, vice president of product marketing at SGI. "We are impressed by how rapidly the Pervasive DataRush platform scaled to tap into the power of Altix UV. Together, they put previously unimaginable power in the hands of bioinformatics, genomic research and other simulation, data mining and mathematical modeling scientists."

The Pervasive DataRush-based Smith Waterman algorithm is available for partners and customers to implement in their environments.