A SDN-based Scheduler For Heterogeneous Cloud Computing Platform

Along with the high-speed development of Internet, the cloud computing technology has developed rapidly. The core idea of cloud computing is a large amount of computing resources together, unified scheduling and management, constitute a cloud computing resources pool provide service to the user. With kinds of business placed to the cloud server, more and more servers added to the data center, resulting in heterogeneous state of new old server performance. At the same time, with the rapid growth of the cloud server cluster traffic and data traffic, computing resources has become a bottleneck of cloud services. In the scheduling management of these limited computing resources inevitably have virtual machine migration, network transmission and other additional overhead. How to make full use of computing resource in the cluster flexibly and efficiently has become an urgent problem.This paper designs a hot data identification algorithm HDcat for Heterogeneous Cloud Computing platform which is accurate and efficient. We use this hot data identification and SDN technology to design a data center optimization scheme. On the premise of a given energy consumption model and topology information, saves the traffic load, reduce the energy consumption of data center, realize the data center of cost optimization and performance improvement. In this paper, we develop a SDN controller of our optimization in the POX platform based on the OpenFlow protocol. Finally we used the API provided by Mininet platform to create the experimental network topology environment, real-time playback experiment is carried out and the experimental results are analyzed in detail using traces on YouTube and WorldCup website, to verify the feasibility of the optimization scheme.