Distributed approach of load balancing in dynamic grid computing environment

摘要

Workload management and resource management are two key aspects in grid computing to provide the better services to grid environment users. Grid computing involves a number of challenges like heterogeneous nature of resources, large number of computing elements, independency of computing resources, different processing capacities of the nodes, different load conditions, overloading of jobs at computing nodes and underutilised resources, etc. In this paper, a distributed grid and dynamic load balancing algorithm is proposed using the forest-based model of the grid structure. The problem of heterogeneity of the resources (the individual node is independent from the physical architecture of a grid) has also been taken into account. In the proposed approach all the computing nodes are connected with a group manager in a group and request to group manager for resource allocation. The group manger is responsible for balancing the workload within the group. In case of non-availability of the resources within the group, the group manger communicates with the other adjacent groups. It has been shown that proposed algorithm firstly balance the load at local level and then goes to the grid level, which results in reduced response time and communication cost.