Modeling and performance evaluation of Smith-Waterman algorithm

Smith-Waterman algorithm is primarily used in DNA and protein sequencing which helps by a local sequence alignment to determine similarities between biomolecule sequences. However the inefficiency in performance of this algorithm limits its applications in the real world. In this perspective, this work presents two fold contributions. It develops and evaluates a mathematical performance model for the algorithm by targeting a distributed processing system. This mathematical model can be helpful to estimate performance of the algorithm for larger size of sequences aligned by the thread level parallelism, using large set of processors configured as distributed processing nodes. Secondly, This work also evaluates in detail the performance scalability of smith-waterman algorithm using OpenMP, MP and a Hybrid (OpenMP + MPI) parallel programming models on a real supercomputing platform Altix-4700. This evaluation shows that the hybrid approach performs better than the other simple approaches.