In recent years, the static and the dynamic jobs scheduling onto heterogeneous processors present a very well studied problem. Typically the Data Grid Scheduling problem (DGS) has recently become an active research area. The heterogeneous processors scheduling problem (HPSP) can be formulated in several ways and the efficient scheduling of the HPSP on the available resources is one of the key factors for achieving high performance results. Historically, finding an optimal schedule was an NP-hard problem in practical cases; researchers have resorted to devising efficient Heuristics and methods inspired by Nature’s Laws. Moreover, the multi-objective scheduling research derives its importance from the need to address the real world of the heterogeneous processors application, which rarely has a single objective function. A schedule that is of a high-quality for one objective function may in fact be quite insignificant for another. Decision makers must carefully evaluate the compromise involved in considering several different criteria in practical scheduling applications. In this paper, we introduce a new hybrid approach that combines ant system optimisation and fuzzy logic concept to facilitate the multi-objective HPSP optimisation, such as the makspean, and the processors workload. Based on the concept of the ant system and fuzzy controller, we automatically control the ant system parameters evolution for the multi-objective HPSP optimisation. The simulation results indicate that the combination of the ant system approach and the fuzzy controller is not only an efficient metaheuristic tool when we search a multi-objective schedules under constraints but also significantly surpasses other scheduling approaches in terms of quality and solution cost.
ant system, fuzzy logic controller, job scheduling, heterogeneous processors, computational grid, large size instances.
All articles published by Studies in Informatics and Control (SIC) are published in accordance with the terms and conditions of the Creative Commons Attribution License (CC BY-NC-SA 4.0).