Accelerating HMMER search using FPGA Grid

HMMER is one of the most used software tools for sensitive profile HMM searches of biological sequence databases. HMMER can be accelerated by comparing a profile HMM with sequences in parallel using a CPU grid. HMMER is, however, still a cpu-intensive program. The search on each node on the grid can be further accelerated by using dedicated hardware. In this paper, we describe an approach for accelerating HMMER search using an FPGA grid. In our system, two prototype circuits which share the same basic processing units are prepared in advance, and the circuit for each size FPGA is generated from the prototype circuits. In the FPGA grid, the difference of the processing speed and data transfer speed becomes much larger than CPU grids. In order to maintain the load balance on each node, new load balancing methods are introduced and evaluated.