An Adaptive Task Scheduling System Based on Real-time Clustering and NetFlow Prediction

With more and more enterprise users begin to adapt cloud computing service as their target platform for important business activity, the Cloud Computing Platform (CCP) is facing severe loading and stability problems. As a result, an excellent task scheduling system is much needed. To build such a system, the widely used method is to design detailed plan manually based on each task. However, this method has many disadvantages. Firstly, people have to add new task and rearrange them every time new user comes, with too much time wasted. Secondly, changing environment like time, task numbers may also influence the final running results. In this paper, we creatively introduce an adaptive task scheduling system which is driven by history user data [1]. By collecting these data, we conduct clustering at intervals of time to achieve the latest user task size classification. We also predict the netflow amount for next few hours, based on historical time series. Later, we arrange the order of current tasks dynamically through a weighted scoring mechanism. Experiments result on the real-life dataset demonstrate the superiority of our proposed method over state-of-the-art method.