Load Balancing: An Optimal Control Problem

One of the primary goals of distributed systems is effective coordinated use of their resources. In a distributed system, an imbalance load distribution on its nodes may reduce the performance of the overall system. Load balancing is a strategy to allocate arriving tasks to the proper nodes to improve the overall performance of the system. This paper introduces a classification for load balancing algorithms and investigates the necessity of using intelligence in these algorithms. A fuzzy logic controller is proposed to enhance the efficiency of load balancing algorithms. The proposed algorithm uses last completed task waiting time as one of the inputs to fuzzy logic controller. The proposed algorithm invokes a process that is called biasing to dispatch arrival load on nodes. The intervals of invoking biasing process are semi-dynamically configured according to traffic intensity of network. Comprehensive simulation shows that the proposed approach is more effective than non-intelligent algorithms.