RC-BLAST: Towards an Open Source, Portable FPGA-based Implementation of BLAST

Colloq: Speaker: 
Ron Sass
Colloq: Speaker Institution: 
Information and Telecommunication Technology Center (ITTC), University of Kansas
Colloq: Date and Time: 
Wed, 2005-06-01 10:00
Colloq: Location: 
ORNL, Bldg. 5700, Room L202
Colloq: Host: 
Jeffrey S. Vetter
Colloq: Host Email: 
vetter@ornl.gov
Colloq: Abstract: 
Basic Local Alignment Search Tool (BLAST) is a common computer application that molecular biologists use to search for sequence similarity in genomic and other biological databases. Researchers in the field of Reconfigurable Computing (RC) have long recognized the potential of using Field-Programmable Gate Arrays (FPGAs) to accelerate sequence matching in general and there have been several demonstrations of this. However, in practice, relatively few molecular biologists use RC technology. Why? One reason is that the FPGA-based solutions to date have implemented custom sequence matching algorithms (that is, not BLAST) and only provide a partial solution. A second reason may be economics: previously low volumes and high costs have made FPGA-based solutions expensive. However, basic FPGA costs are dramatically less expensive today (and are decreasing much faster than conventional computer hardware). Also, the number of vendors selling off-the-shelf FPGA hardware increasing. Finally, several traditional High-Performance Computing companies are including FPGA options in their latest offerings. Hence, now is an opportune time to re-evaluate the feasibility of accelerating BLAST with FPGAs. Since BLAST is such a common tool, it is possible that a more cost-effective BLAST solution may enable scientific discovery that currently is impeded by cost or computation time. The goal of the RC-BLAST project is to develop a portable, Open Source FPGA-based implementation of NCBI's BLAST. This talk describes our experiences developing RC-BLAST over several generations of FPGA-based reconfigurable computing cards. The talk details the key aspects of our most recent design, currently running on an AceIIcard. The portability and cost-effectiveness of this system are evaluated and, based on these results, we present lessons learned. This is an on-going effort and the talk concludes with a discussion of the next RC-BLAST design in progress.
Colloq: Speaker Bio: 
N/A