A Framework of Using Mobile Agent to Achieve Efficient Load Balancing in Cluster

In a cluster of computers, if the system load in each of computers is nearly equal, it indicates good resource utilization. It is well-known that load balancing is a key factor in developing parallel and distributed applications. In our paper, a mobile agent based load balancing technique is presented. Instead of balancing the load in cluster by process migration, or by moving an entire process to a less loaded computer, we make an attempt to balance load by splitting up processes into separated jobs and then balance them to nodes by our novel load balancing method. In order to get our target, we use mobile agent to distribute load among nodes in a cluster. We have created five kinds of agents in namely directory agent, account agent, threshold agent, monitor agent and client agent. These agents follow the required policies for load balancing. Our proposed system focuses on problems that are inherently parallel or problems that can be worked with reasonable amount of synchronization. Our load balancing protocol is being implemented using Aglettrade, a Java_compliant mobile agent platform from IBM. Although we cannot claim that a mobile agent enable approach will always outperform message passing based load balancing, such an alternative approach to designing load balancing protocols has several advantages. Preliminary experimental results demonstrated that the proposed technique is effective