基于流量特征的流调度策略研究综述

软件定义网络将网络的数据层和控制层相分离，具有可编程性强和全局网络视图的优点，这一优点被越来越多的应用于数据中心网络流调度中。然而，数据中心网络中大象流和老鼠流共存的流量特征是流调度技术的一大挑战。对基于流量特征的流调度策略进行了综述，首先梳理了基于软件定义网络的数据中心网络的理论基础，然后对大象流、老鼠流、大象流与老鼠流三个方面的流调度策略进行了分析和总结，并探讨了相比于传统的流调度技术，在软件定义网络架构下基于流量特征的流调度策略的优势。最后针对目前基于流量特征的流调度技术面临的若干主要问题进行了讨论，指出了下一步的研究方向。

Survey of research on flow scheduling strategies based on traffic characteristics

Software defined network proposes to decouple the control and data forwarding plane, with the advantages of strong programmability and a global network view. This advantage is increasingly used in data center network flow scheduling. However, the traffic characteristics of the elephant flow and the mice flow coexisting in the data center network are a major challenge in the flow scheduling technology. This paper reviewed the traffic scheduling strategy based on traffic characteristics. Firstly, it investigated the theoretical foundation of data center network based on software-defined network. Secondly, it analyzed and summarized the flow scheduling strategies of elephant flow, mice flow, elephant flow and mice flow, and discussed the advantages of the flow scheduling strategy based on traffic characteristics in the software-defined network architecture compared with the traditional flow scheduling technology. Finally, this paper discussed the potential challenges in flow scheduling based on traffic characteristics and direction of future research.