The Artificial Bee Colony Algorithm Applied to a Self-adaptive Grid Resources Selection Model

Swarm intelligence algorithms are used to simulate the behaviour of non-centralized and self-organizing systems, which could be natural or artificial. Grid computing environments are distributed systems comprised heterogeneous and geographically distributed resources. This computing paradigm presents problems related to resources management (discovery, monitoring and selection processes) which are caused by its dynamic and changing nature. These problems lead to a bad application performance due to the fact that resources availability and characteristics vary over time. In recent years, several approaches based on adaptation and defined from a system point of view have been proposed. The present contribution is focussed on enhancing the grid resources selection process by providing a self-adaptive ability to grid applications. A selection model based on the Artificial Bee Colony algorithm is described. In contrast to other alternatives, the model is defined from a user point of view (the model has not control on the internal grid components). Finally, the approach is tested in a real European grid infrastructure. The results show that both a reduction in execution time and an increase in the successfully completed tasks rate are achieved.