Parallel Sequence Analysis under Three Load Balancing Strategies

Retrieving homologous sequences from existing databases is very important for the biomedical research community. In this project, we implemented four sequence alignment algorithms (general gap penalty, affine gap penalty, general gap penalty with mid-point and affine gap penalty with midpoint) to look for nearly optimum aligned sequences in the database for nucleic acid sequences (i.e. DNA/RNA sequences). We use MPI, which is a MIMD programming interface, to achieve parallelism. Our focus in this project is to achieve good load balancing strategies. Three load balancing strategies, including static assignment, dynamic assignment and semi-dynamic assignment, are implemented to divide and assign sequences in the database to different processors. And the semi-dynamic assignment strategy is the highlight. It is designed by us to achieve better load balancing than pure static and dynamic load balancing strategies. We also compare the performance of these three load balancing strategies.