An Agent-Based Migration Transparency and Fault Tolerance in Computational Grid

A Grid is a large-scale, geographically distributed hardware and software infrastructure for flexible, secure, and coordinated sharing of vast amounts of heterogeneous resources within large, dynamic and distributed communities of users belonging to virtual organizations, to enable solving of complex scientific problems. From the perspective of one computer, such network partitioning may appear as a failure to other computers. These types of failures may lead to major impact on whole application which is executing on Grid for many days. Thus, failures in the grid computing environment can be solved to some extent by performing migration of application through node agents. As executing applications in nodes have contiguous service features, it becomes important to handle and mask fault and migrate the current job to another grid node without stopping the on-going processes. In this paper, we use agents to provide communication between grid nodes and handle failure tolerant techniques on grid. The agent is dynamically reallocated to Grid nodes though a transparent migration mechanism, as a way to provide fault tolerance in computational grids