Adaptive workflow scheduling scheme based on the Colored petri-net model in cloud

Abstract

In the science area, workflow management systems are used for the purpose of managing collaborative researches of many research organizations, utilizing the distributed computing resources effectively, etc. In addition to this, as the cloud computing technology being introduced, many researches on enhancing the economic feasibility or the efficiency of workflow management systems are done. While executing a workflow application, a workflow management system should decide the resource to assign each task and the order to execute the assigned tasks. The processing cost and the completion time can be very different depending on that schedule. Furthermore, users can ask for low processing cost or short completion time, but satisfying those two requests at the same time is practically impossible. Therefore, the existing workflow scheduling schemes try to find the optimal solution while setting bounds to one condition. These schemes can find some schedules that satisfy their own purposes. However, they cannot get the schedules satisfying all the various SLAs that vary depending on users, workflow applications, and so on. Therefore, we propose the adaptive workflow scheduling scheme based on the colored Petri-Net model which offers two selective workflow scheduling policies. The proposed scheme separates the scheduling phase and the execution phase, and distributes the actual remaining time or remaining budget in the execution phase in the ratio of the processing time or processing cost assigning rate which is decided in the scheduling phase. We defined the procedures of each phase using the colored Petri-Net model. As a result of our experiment, we can get the right schedules that fulfill their own purpose according to the selection of a scheduling policy.