Improving extremal optimization in load balancing by local search

The paper concerns the use of Extremal Optimization (EO) technique in dynamic load balancing for optimized execution of distributed programs. EO approach is used to periodically detect the best candidates for task migration leading to balanced execution. To improve the quality of load balancing and decrease time complexity of the algorithms, we have improved EO by a local search of the best computing node to receive migrating tasks. The improved guided EO algorithm assumes a two-step stochastic selection based on two separate fitness functions. The functions are based on specific program models which estimate relations between the programs and the executive hardware. The proposed load balancing algorithm is compared against a standard EO-based algorithm with random placement of migrated tasks and a classic genetic algorithm. The algorithm is assessed by experiments with simulated load balancing of distributed program graphs and analysis of the outcome of the discussed approaches.