SDN环境跨层P2P流量优化方法研究

Currently P2P applications occupy a large amount of Internet bandwidth. SDN has centralized control, open northbound interface, programmable networking and other features, and provides support for emerging or core network technologies such as segment routing and network coding, which brings new solutions to P2P traffic optimization problems. This project first establishes the traffic engineering model of the transmission process of the P2P system in the SDN environment. On this basis, the transmission characteristics of periodicity, dynamics and discretization of unstructured P2P systems are deeply explored. Two modes of cross-layer optimization methods are studied in depth: The unicast mode uses segment routing, where the intermediate nodes are optimally time-division multiplexed to consecutive P2P transmission periods using Markov approximation theory; The multicast mode uses network coding, where the unicast transmission in units of nodes is converted into network coded multicast transmission in units of media chunks. Finally, using network flow optimization theory to improve scalability, reduce computational overhead, and adapt to dynamics, the network layer optimization algorithms that can be deployed in SDN are studied, including decomposition algorithm, approximation algorithm and online algorithm. The project innovatively combines stochastic optimization and network flow optimization theories. The research results can reduce the P2P traffic occupation and improve the utilization efficiency of network resources while ensuring the performance of P2P system. It has both theoretical innovation significance and practical application value.

当前P2P应用占据了大量互联网带宽。SDN具有集中控制、北向接口开放、网络可编程等特性，并支持分段路由和网络编码等新兴或核心网络技术，为P2P流量优化问题带来了新的思路。本项目首先建立SDN环境P2P系统传输过程模型。在此基础上，深入挖掘了P2P系统周期性、动态性、离散化等传输特征，重点研究两种模式的跨层优化方法：单播模式采用分段路由，利用Markov近似理论将分段路由中间节点时分复用至P2P传输周期；组播模式采用网络编码，将P2P传输周期内的以节点为单位的单播传输转化为以片段为单位的网络编码组播传输。最后，利用网络流优化理论，以提高可扩展性、降低计算开销、适配动态性为目标研究可部署于SDN的网络层优化算法：分解算法、近似算法和在线算法。本项目创新性地融合了随机优化和网络流优化理论，提出的优化方法能够在保障P2P系统性能的同时降低P2P流量占用、提升网络资源利用率，具有理论和应用价值。

流媒体传输与文件共享类应用已覆盖了绝大多数互联网用户并占用了绝大部分互联网流量，与之同时，迅速增长的网络流量需求与有限的网络资源之间的矛盾困扰着网络运营商和用户。本项目在SDN环境下借助分段路由和网络编码技术，提出一套适用于应用层和网络层跨层流量优化的解决方案，可广泛支持单播、组播、服务功能链、P2P等流量类型，具体包括：1）针对SDN环境单播分段路由流量优化问题，提出可扩展分段路由优化框架与优化算法，所提出算法是目前唯一针对任意分段路由流量工程问题的多项式复杂度求解算法。作为分段路由的延伸应用，针对无线Mesh网络吞吐量优化问题，提出分段路由辅助的无线Mesh网络架构和具有理论性能保证的优化算法；2）针对会话内网络编码多源组播优化问题，基于极流分解方法建立了全新的计算模型，并提出简洁、高效、可扩展，并具有理论性能保证的近似算法和在线算法；3）针对基础的在线路由问题，提出在线多路径路由乘子权更新算法，所提出算法在不同网络类型中显著改善了路由性能，并且在理论上被证明在不违反网络链路带宽限制的前提下具有对数级别的竞争比；4）针对应用层和网络层跨层优化问题，提出了应用层和网络层跨层优化框架与快速收敛算法，能够解决应用层和网络层优化目标不一致和信息不同步导致的震荡问题，并具备应用层和网络层错误恢复能力。