The design, implementation, and experimentation of a grid-enabled framework for metaheuristics, ParadisEO-CMW, are dealt with in this paper. The basic components of this new framework are the known open-source software tools EO (Evolving Objects), which allows for the flexible design of evolutionary algorithms; ParadisEO (Parallel and Distributed EO, previously introduced by the same authors), which provides the most common models for parallel and distributed computing; Condor, which deals with heterogeneous computing resources; and MW (Master-Worker), which allows for easy creation of master-worker type parallel applications on Condor-enabled infrastructures. It should be mentioned that the proposed framework is one of the first attempts to port metaheuristics applications to grid environments with positive results.

The experiment proving the framework’s usefulness was performed on a near-infrared spectroscopic data mining application intended to discover a predictive model for the concentration of sugar in beets. The results were obtained using a large campus cluster of desktop computers, and show that the new proposed framework efficiently exploits the available resources.

The content of the paper is useful, especially for metaheuristics system designers. The authors pay special attention to technical details related to the new framework and experimentation results. A nice introductory presentation of the parallel models for evolutionary algorithms is also provided.