Fast Multi-objective Rescheduling of Workflows to Constrained Resources Using Heuristics and Memetic Evolution

Abstract

Scheduling of jobs organized in workflows to a computational grid is a permanent process due to the dynamic nature of the grid and the frequent arrival of new jobs. Thus, a permanent rescheduling of already planned and new jobs must be performed. This paper will continue and extend previous work, which focused on the tuning of our G lobal O ptimising R esource B roker and A llocator GORBA in a static planning environment. A formal definition of the scheduling problem and a classification will be given. New heuristics for rescheduling based on the old plan will be introduced and it will be investigated how they contribute to the overall planning process. As an extension to the work published in Conf. Proc. PPAM 2009, LNCS 6067 or 6068 (to be published in July, 2010) a simple local search is added to the basic Evolutionary Algorithm (EA) of GORBA and it is examined, whether and how the resulting Memetic Algorithm improves the results within the limited time frame of three minutes available for planning. Furthermore, the maximal possible load, which can be handled within the given planning time, will be examined for a grid of growing size of up to 7000 grid jobs and 700 resources.