Cost optimization based on brain storming for grid scheduling
María Arsuaga-Ríos, Miguel Angel Vega-Rodríguez · 2014
Job scheduling is a challenging task in terms of execution time, which is usually minimised to accomplish dead-lines in diferent areas. However, cost optimization is a conflictive objective when the execution time is optimized, because usually expensive resources are the fastest ones. In this paper, we present a multiobjective algorithm based on the brain storming process that human groups carried out in order to achieve the best initiative or ideas to optimize the job scheduling problem in Grid environments. This swarm algorithm, MOBSA (multiobjective Brain Storm Algorithm), minimizes both objectives: execution time and cost without any weights of influence. The standard multiobjective algorithm NSGA-II is compared with MOBSA to evaluate its multiobjective quality. The results obtained by MOBSA are superior than the results obtained by MOHEFT which is the multiobjective version of HEFT, one of the most-used algorithms in workflow scheduling. Furthermore, MOBSA's results also surpass the results provided by real grid schedulers as WMS from gLite middleware or DBC from Nimrod-G.