Load balancing framework for distributed systems

Development of efficient load balancing methods that are sensitive both to static and dynamic loads, in distributed systems environments remains to be challenging. Static load balancing is applied to cases where the application and system requirements are known before the execution. Dynamic load balancing, on the other hand, is applied to systems where the load on individual processors varies, during the time of execution. In this paper, an integrated load balancing approach is presented to deal with sporadic or periodic load changes in the system. The algorithm is activated when the load changes reach a threshold, since the last activation. The approach integrates the benefits of both static and dynamic load balancing. The module relocations are taken into consideration while application advances from phase to phase, during the execution. An n-phase problem is first reduced to a two-phase problem. It is then solved using a modular heuristic approach. The results obtained are very close to those of optimum solutions.The framework allows implementation flexibility where clustering, mapping, and migration components of the algorithm can independently be changed.