Computational Load Management in Grid Application – A Dynamic Perspective

Abstract

This paper focuses on experiences gained during the development and implementation of dynamic load balancing algorithm required for efficient handling of compute intensive application and also revelation on an agent based approach for handling interactive jobs in the grid environment. A computational Grid is a hardware and software infrastructure that provides dependable, consistent, pervasive and inexpensive access to high end computational capabilities. To acquire effective load balancing in an environment where workstations typically operate under continuously changing conditions, extended load predict for computational resources is increasingly required. This algorithm dynamically distributes the computational load among the available grid nodes according to its availability. Many innovative schemes for allocating and handling jobs in a grid computing system have been proposed in order to achieve highly utilized distributed and parallel computing system. Here we propose an agent based approach for managing interactive jobs, where agent is a software entity showing several degrees of sovereignty. In an interactive jobs, nodes are reserved for users to log into and issue commands by hand, runs immediately upon submit which never wait. Although interactive job may decrease the utilization of the system, it is vital for a working scientific computing system.