软件定义网络的测量方法研究

测量技术是状态监测、性能管理、安全防御等网络研究的基础，在网络研究领域具有重要地位.相较于传统网络，软件定义网络在标准性、开放性、透明性等方面的优势给网络测量研究带来了新的机遇.测量数据平面和测量控制平面的分离，启发了通用和灵活的测量架构的设计与实现；标准化的编程接口，使得测量任务可以快速地开发和部署，中心化的网络控制可以基于反馈的测量结果实时地优化数据平面的硬件配置和转发策略，数据平面基于流表规则的处理机制支持对流量更加精细化地测量.但是，软件定义网络测量中额外部署的测量机制造成的资源开销与网络中有限的计算资源、存储资源、带宽资源产生了矛盾，中心化的控制平面也存在一定的性能瓶颈，这是软件定义网络测量研究中的主要问题和挑战.分别从测量架构、测量对象两方面对当前软件定义网络测量研究成果进行了归纳和分析，总结了软件定义网络测量的主要研究问题.最后，基于现有研究成果讨论了未来的研究趋势.

Survey on Measurement Methods in Software-defined Networking

Measurement technology plays an important role in the field of network researching as it is the foundation of researching on network monitoring, network management, and network security. Compared with the legacy network, software-defined network brings new research opportunities due to its advantages of standardization, openness, and transparency. The separation of measurement data plane and measurement control plane motivates the design and realization of more general and flexible measurement frameworks. Standardized API enables quick development and deployment of measurement tasks. Logically centralized controller enables the network to optimize the hardware configuration and forwarding strategy in real time by analyzing the measurement results pulled from SDN switches. The forwarding mechanism based on flow table enables data plane to provide more fine-grained measurement data of network traffic. However, challenges are also brought due to the contradiction between extra network overhead and the limited computing, memory, bandwidth resources in network. The centralized control plane may also suffer from performance bottlenecks. This paper reviews current researches on measurement methods of software defined networking from two aspects:framework and object, and then concludes the main research tasks and challenges. In the end, the argument about future research trend is elaborated based on existing research experiences.