|
VT and RNET Receive NSF Grant for Genome Sequencing with a SmartNIC |
|
|
|
 Wu Feng of the Departments of Computer Science (CS) and Electrical and Computer Engineering (ECE) at Virginia Tech, and Gerald Sabin (photo, left) and V. “Nagu” Nagarajan (photo, right) of RNET Technologies recently received a National Science Foundation Small Business Technology Transfer (STTR) grant of $150,000 for the first phase of a research project entitled “Network Offloading for Genome Sequence Searching using the SmartNIC.”
In the past decade, there has been an increasing research focus on computational biology as a means to provide sophisticated mechanisms to study and understand the composition and functional capabilities of biological entities and processes. The increasing sizes and capabilities of today’s high-end computing systems provide the ability to interact with and analyze existing information about well known biological entities and utilize it to improve insights into the behavior of newer entities.
Given the importance of sequence search, researchers have designed a number of tools to perform them in an efficient manner. Among the most widely used sequence-search tools is the Basic Local Alignment Search Tool (BLAST) from the National Center for Biotechnology Information (NCBI). BLAST is a multi-threaded but sequential tool that identifies regions of local similarity between sequences by calculating the statistical significance of local matches. To accelerate sequence search, Feng’s Synergy Lab at Virginia Tech (VT) created mpiBLAST, a freely available and open-source parallelization of BLAST using the MPI communication library that delivers super-linear performance speed-up compared to NCBI BLAST.
With the size of the GenBank sequence databases doubling every 12 months and computational horsepower doubling only every 18-24 months, the growth rate of the databases has been fast outpacing the performance of general-purpose processors. Accordingly, scaling parallel sequence-search mechanisms such as mpiBLAST to next-generation large-scale systems is becoming increasingly important in order to meet the growing processing requirements of parallel sequence search.
Thus, the current STTR grant logically extends the work done by Dr. Feng’s lab in order to design a Field Programmable Gate Array (FPGA) based co-processing unit (and its associated software) that offloads some of the prime data-management functionality of the sequence-search tools and applications such as mpiBLAST, amongst others. This co-processing unit, termed as the Sequence-Search Data Management SmartNIC is intended to interact with both the host system (hosting the sequence databases and the core of the sequence search software stack) as well as the existing network interconnects (e.g., InfiniBand, 10-Gigabit Ethernet) to not only utilize their capabilities, but also to supplement them with more advanced communication as well as application relevant features that are not available today.
Together with the SmartNIC hardware, the current grant also intends to provide a software emulation of the SmartNIC, that allows one core in a multi-core system to be used as a co-processing unit providing similar functionalities as the SmartNIC. Though the software emulation is not expected to provide all the features that the SmartNIC hardware can, it can be still be used as a low-cost option in some environments. |
|
