A genetic approach to dynamic load balancing in a distributed computing system

Presents an efficient dynamic load balancing scheme based on a genetic algorithm (GA) which includes an evaluation mechanism of fitness values in a changing environment. Sender-initiated task migration algorithms continue to send unnecessary requests for a task migration while the system load is heavy, which yields inefficient inter-processor communication and much overhead until the migration is actually performed. In the proposed GA-based load balancing scheme, a subset of processors to which the requests are sent is adaptively determined by a learning procedure to reduce unnecessary requests. The learning procedure consists of standard genetic operations, such as selection, crossover and mutation, applied to a population of binary strings, each of which stands for a list of processors to which the migration requests are sent. Each processor has its own population, and the fitness of a string depends on how efficiently the destination of a migration is found. From the viewpoint of the mean response time of the whole system, we show the effectiveness of our approach through empirical investigations