Evaluation of Particle Swarm Optimization Applied to Grid Scheduling

The problem of scheduling independent users' jobs to resources in Grid Computing systems is of paramount importance. This problem is known to be NP-hard, and many techniques have been proposed to solve it, such as heuristics, genetic algorithms (GA), and, more recently, particle swarm optimization (PSO). This article aims to use PSO to solve grid scheduling problems, and compare it with other techniques. It is shown that many often-overlooked implementation details can have a huge impact on the performance of the method. In addition, experiments also show that the PSO has a tendency to stagnate around local minima in high-dimensional input problems. Therefore, this work also proposes a novel hybrid PSO-GA method that aims to increase swarm diversity when a stagnation condition is detected. The method is evaluated and compared with other PSO formulations, the results show that the new method can successfully improve the scheduling solution.