Using system state information for adaptive state polling policy in distributed load balancing

The location policy in distributed load balancing schemes locates the destination nodes to or from which tasks will be transferred. It should evenly distribute workload to the entire nodes with minimal delay for transferring task. The traditional policies can be classified into dynamic selection, random selection, and state polling. However, the policies representatively cause unpredictable state, excessive task transfers, and useless polling problems, respectively. An efficient adaptive location policy is required in the sense that it can react to changes in system state and achieve high performance. We propose on advanced state polling policy based on predictable system state information. The system state information is composed of the state information collected at run time and the predefined static information that is a global priority order of each node for transferring tasks. The global priority order is generated by the global priority network. When load balancing is triggered at a heavily loaded node, the proposed location policy dynamically predicts lightly loaded nodes and other heavily loaded ones by exploiting predictable state information. Then it adaptively finds a good lightly loaded node that minimizes useless polling and maximizes even load distribution. An analytic model is developed to compare the presented policy with other well known policies. The validity of the model is checked with an event driven simulation, and it is shown that the proposed policy exhibits a significant performance improvement over other policies.