Two Hierarchical Dynamic Load Balancing Algorithms in Distributed Systems

In this paper two new methods for load balancing in distributed systems are proposed. Both methods are based on hierarchical structure. Hierarchical structure provides better load management because it allows algorithms to balance the loads in two levels of groups and nodes. Since they have centralized load balancing mechanism, they have lower communication overheads too. These methods are dynamic and have simple implementation. The first method use biasing process to allocate weights called biases. Biases are determined based on the current load state of the groups and nodes. Second method improves round robin algorithm in a way that the group or node with minimum load state have priority on others to receive tasks from load balancers in a specific time slice. Comparative study of these methods is done for both homogeneous and heterogeneous nodes and it shows better performance than existing conventional algorithms.