"Using the Starbridge Systems FPGA-based Hypercomputer for Cancer Research"
Jack R. Collins1, Kent Gilson2 and Jim Yardley2
1 National Cancer Institute/SAIC Frederick
2 Starbridge SystemsAbstract
The computational demand required to process and annotate biological data resulting from high-throughput technologies far outstrips the ability of traditional computing to meet the challenge of converting the data into information or knowledge in a timely manner. Simply annotating and updating current databases has become a task that occupies large fractions of many groups’ computational resources in order to complete their task before the next release of the data. Therefore, computational resources must be employed that exploit as much concurrency as possible within the mathematical model in an attempt to increase the compute density of a given architecture. We have investigated new computational technologies based on FPGAs with the goal of providing newer and more powerful tools to the biological community. We have collaborated with Starbridge Systems (SBS) implement a Smith-Waterman algorithm on an SBS Hypercomputer using their Viva programming language. Using Viva to create a pipeline structure, we have created an adaptable software system that can address all of the FPGAs within the system. Currently, the sequence is continuously read into the system as the SW matrix is being generated. We discuss and test the scalability and efficiency of the system using several different biological sequences with the largest comparing the human X and Y chromosomes, 147 million and 60 million bases respectively. Insights and experiences using Viva and the SBS hypercomputer will also be presented.