Summary: BLAST is one of the most popular computational biology tools. The execution cost of BLAST is highly dependent on database sizes, which have considerably increased following all recent advances in sequencing methods. The evaluation of BLAST in distributed and parallel environments like PC clusters and Grids has been largely investigated in order to obtain better performances. This work evaluates a replicated allocation of the (sequences) database, where each copy is also physically fragmented. We investigate two dynamic workload balancing methods that focus on our database allocation strategy. Preliminary practical results show that we achieve both a balanced workload and very good performances. We briefly discuss ideas that would make our approach feasible for Grid computational environments.