An adaptive load balancing algorithm with use of cellular automata for computational grid systems

Abstract

Load balancing algorithms play a challenging, complicated, and important role in the performance of computational Grid systems. In this paper, we present a decentralized adaptive load balancing algorithm with use of cellular automata, named LBA-CA. Each computing node in the Grid system is modeled as a cell of proposed cellular automata and can be in four states. Cellular automata (abbreviated to CA) are used for designing a load balancing algorithm for computational Grids because of its distributed and dynamic manner. In addition, such natural properties of CA make LBA-CA an appropriate local load balancing algorithm for each cluster of computational Grids. Due to resource heterogeneity and communication overheads exist in computational Grid systems; we take account of several issues in LBA-CA such as processing power of computing nodes and communication latency. The main goal of our algorithm is to reduce the average response time of arrival jobs. The performance of our algorithm is evaluated in terms of several metrics including the average response time of jobs, processor utilization, percent of executed jobs, and average Off time in relation to considerable variations in transition time, service time, and number of jobs. © 2011 Springer-Verlag.