BLAST Distributed Execution on Partitioned Databases with Primary Fragments
Daniel Xavier de Sousa (PUC-Rio Depto Informatica, Brazil)Sergio Lifschitz (PUC-Rio Depto Informatica, Brazil)
Patrick Valduriez (INRIA and LINA, Nantes, France)
Abstract:
As a result of recent advances in sequencing methods, genomic databases are getting larger and larger, thus raising performance issues for bio-sequence analysis tools. In this paper, we consider BLAST, one of the most popular such tools. To increase performance in large databases, much work has considered the evaluation of BLAST in distributed and parallel environments like clusters and Grids.
We propose a new parallelization approach to execute BLAST in distributed and parallel environments. We use a replicated allocation of the (sequences) database, where each copy is physically fragmented. We investigate two dynamic load balancing methods that exploit
our database allocation. Our preliminary experimental results based on a cluster indicate that our approach achieves both very good speedup and good load balancing.
Keywords:
Data-intensive Grid applications, Data Grids, Web services and P2P, Replication and fault-tolerance in Grids, Indexing, caching and load balancing in Grids
