Multi Workflow Fair Scheduling Scheme Research Based on Reinforcement Learning

Abstract

In this study, aiming to optimize the multi-workflow scheduling order, in which tasks submitted at different time require different service quality, we present a fair multi-workflow scheduling scheme based on reinforcement learning. Firstly we design a dynamic priority-driven algorithm, in order to set the initial state of the task priority according to the type of cloud workflow and service quality on the one hand, and on the other hand, to adjust the tasks priority dynamically while scheduling so as to avoid violating the Service Level Agreement by delaying the workflow provisioning. Secondly, we design a fine-grained cloud computing model and apply the reinforcement-learning based scheduling algorithm to balance the cluster loads. Finally the experimental results prove the effectiveness of this scheme.