Design of a workflow management system for BLAST servers using Jini and JavaSpaces

BLAST (Basic Local Alignment Search Tool) is an important DNA and protein similarity tool. It can give evolution and functionality information about unknown genes by comparing them with well studied genes in a database. BLAST search may query many DNA or protein sequences against a huge collection of databases. Today, it is normal that one biological laboratory performs BLAST searches for thousands of sequences in its daily research activities. The performance of the BLAST search implementation is a bottleneck of such activities. In this thesis, we present a design of a new distributed BLAST cluster system using Jini and JavaSpaces technology. This system contains a set of BLAST servers. Each server has a partition of a database against which BLAST algorithms may be executed. This thesis focuses on designing a workflow management system which coordinates the work of distributed jobs. Jini and JavaSpaces provide an easy way to exchange program code and data over a network. We found it is easy to implement a workflow management application in a distributed computing environment using Jini technology. Although this system is for BLAST search, we found it is not difficult to modify it to fit other distributed computing tasks.