Dynamic and Adaptive Load Balancing using Harmony Search in Grid Computing

Applications of network-based systems like cluster and grid systems have increased considerably in recent years. Load balancing is an important concept in these systems, which is implemented with the purpose of reducing execution time of applications and increasing utilization of resources. An algorithm is proposed in this paper for load balancing in the grid systems which operates on the basis of harmony search algorithm. This algorithm which is called Dynamic and Adaptive Harmony Search (DAHS) distributes the load among heterogeneous resources in a centralized and dynamic manner. DAHS parameters are defined adaptively in order to solve the difficulties of initialization. For creating a better load balancing, the amount of load in each resource is evaluated after formation of every harmony, such that a new task will not be assigned to it if a resource has extra load. The required simulations are done in GridSim simulator to investigate performance of this algorithm and it was demonstrated that utilization of resources was greater than 92% in DAHS method with mean square deviation being smaller than 0.06%. Meanwhile, makespan of the proposed method was found 10% lower in comparison with that of genetic method.