基于算法原语的SDN数据平面关键技术研究

Programmable data-plane is essential to unleash the power of Software Defined Networking (SDN) in network innovations due to it is capable of forwarding packets without knowing any specific protocols and routing mechanisms. This project aims to scale the performance of SDN data-plane, to ease of data-plane programming on heterogeneous match hardware, and to provide an efficient sharing mechanism between heterogeneous match resources in SDN network. In this project, we present algorithmic primitive based data-plane which views different match tables (TCAM tables, Trie tables, hash tables etc) as algorithmic primitives and combine these primitives to support scalable data-plane; We propose rule-set analysis methods and fast incremental update methods to realize heterogeneous match resources allocation, and propose efficient match resource sharing mechanisms to support multiple network applications(measurement, QoS) running simultaneously. The innovative outcomes of this project will provide new insight, methods and mechanisms for programmable SDN data-planes, which will accelerate the deployment of SDN in production networks.

可编程数据平面能够实现协议无关的数据包转发，是软件定义网络（SDN）简化业务创新的重要前提。本项目针对可编程数据平面中匹配性能伸缩性，异构匹配资源可编程性和多匹配应用(如测量、QoS等)对匹配资源的竞争问题开展研究，提出基于算法原语的数据平面架构，将异构的匹配资源（如Trie表，哈希表，TCAM表）视为算法原语进行组合以实现灵活可伸缩多协议字段匹配;提出规则分析方法和规则快速增量更新方法实现自动分配异构匹配资源，简化编程；提出高效匹配资源共享机制以支持不同的匹配应用。本项目为SDN数据平面的可编程性提供新的方法和机制，对SDN技术在实际中应用具有重要意义。

可编程数据平面能够实现协议无关的数据包转发，是软件定义网络（SDN）简化业务创新的重要前提。本项目针对可编程数据平面中匹配性能伸缩性，异构匹配资源可编程性和多匹配应用(如测量、QoS等)对匹配资源的竞争问题开展研究，提出基于算法原语的数据平面架构，将异构的匹配资源（如Trie表，哈希表，TCAM表）视为算法原语进行组合以实现灵活可伸缩多协议字段匹配;提出规则分析方法和规则快速增量更新方法实现自动分配异构匹配资源，简化编程；提出高效匹配资源共享机制以支持不同的匹配应用。本项目为SDN数据平面的可编程性提供新的方法和机制，对SDN技术在实际中应用具有重要意义。