KMI is a distributed, in-memory hashing library for matching k-mers and short snippets of DNA sequences. The original KMI code was originally implemented using asynchronous MPI communica-tion. In Figure 12 we show how the performance of the original implementation compares to that of the re-implementation of KMI on top of our communication library. We used two systems in the performance evaluation: 32 BG/Q nodes and 32 P7-IH nodes. The results show that the KMI implementation using our model is 9X and 10X faster than the MPI implementation on P7-IH and BG/Q, respectively.