Neighbour Aware Random Sampling (NARS) algorithm for load balancing in Cloud computing

Abstract

Load balancing plays a vital role in Cloud computing to enhance throughput, optimize resource use and reduce response time. The main features to be considered while selecting a load balancing algorithm for cloud is the ability of the algorithm to address distributed network, dynamic environment and self-regulation. Biased random sampling is one such algorithm; it allocates jobs by performing a random walk in the network. The selection of neighbour is uniformly distributed among the neighbour nodes in case of biased random sampling algorithm. Improved version of random sampling algorithm uses cost based load computation to select node for random walk [1]. This paper introduces neighbour awareness and prediction mechanisms to further improve the selection process of nodes for random walk. The proposed algorithm selects the least loaded node from the neighbour list for the random walk. This can be achieved by computing probability of each neighbour based on perceived load of the respective neighbour. Thus the probability of choosing lightly loaded node can be increased and hence the job waiting time can be decreased further.