We are looking to leverage the two key insights that we have for parallelizing computational genomics applications to parallelize other applications beyond our initial target of genomic sequence alignment. We observe that these applications have parallelism in two dimensions— the first is in repeated invocations of the algorithm for matching different “queries” (we call this inter-query parallelism or coarse-grained parallelism (CGP)) and the second is the scope to match different parts of an individual query in parallel (we call this intra-query parallelism or fine-grained parallelism (FGP)).

We are looking to model the performance of these applications at large scales - large process counts plus large data sizes. This will enable us to run them at large scales and detect and diagnose any performance problems automatically.