Hybrid Optimization Techniques for Fuzzy Logic Controller Design in Parallel Job Scheduling Problems

摘要

In this study attempt is taken to improve the performance of the fuzzy logic controller employed infinding solutions for the parallel job scheduling problems. The performance of the fuzzy logic controllerdepends on its knowledge base which consists of data base and the rule base. This paper proposes novelhybrid optimization techniques for performance improvement of the fuzzy logic controller by optimizing itsknowledge base and a comparative analysis of the proposed optimization techniques are presented based onthe computed simulation results. Scheduling of parallel jobs is one of the most challenging aspects with respectto analyzing the performance of the parallel system process. In a parallel system, if the application containsprocesses which are not co-scheduled together, then the performance of the parallel system starts degrading.Agile Scheduling algorithm classifies the grain sizes in a detailed manner for the real workloads and schedulesthem in an effective manner. Using the results obtained from the agile scheduling algorithm, a rule based systemis generated which classifies all the scheduling states and assigns the appropriate scheduling class for theparallel jobs. The rule system is coded with the Mamdani Fuzzy model and to improve the modeled Fuzzy LogicController (FLC), the proposed optimization techniques are applied over the knowledge base of the fuzzy logiccontroller which involves optimization of both the database and rule base simultaneously. This paper employsoptimization algorithms like Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Ant ColonyOptimization (ACO) and River Formation Dynamics (RFD) for fuzzy logic controller design of parallel systemsand as well hybrid technique along with the tabu search algorithm. Simulation results prove the effectivenessof the developed algorithms for fuzzy logic controller design of parallel job shop scheduling problems.