Tuning of Operators in Memetic Algorithms for Independent Batch Scheduling in Computational Grids

Efficient scheduling of jobs to resources is a core service of Computational Grids (CGs). Due to the large scale, the dynamic nature and the highly heterogeneous tasks and resources, scheduling is a challenging problem in CGs. Different methods, from simple heuristics to more sophisticated optimization and artificial intelligence techniques, have been for a while now in the research and development agenda of researchers of the Grid computing community. One family of algorithms that represent interest is that of Memetic Algorithms (MAs), a variant of evolutionary algorithms that combines genetic search with local search. In this paper we present a study on the tuning of the operators in MAs for the problem of Independent Batch Scheduling in Computational Grids. The aim is to identify a combination of operators and parameters that would lead to the design of robust Grid schedulers using MA solvers that compute high quality planning in very short times. For the study, we have used both a static benchmark of instances and a Grid simulator to capture realistic features of real Computational Grids.