REFINE:一种可重构的SDN转发平面实现模型

发布时间：2018-06-19 13:40

本文选题：可重构 + SDN转发平面抽象　；参考：《小型微型计算机系统》2015年10期

【摘要】：SDN(Software-Defined Networking)转发平面抽象是有效提升网络控制能力的关键.由于Open Flow描述能力有限,难以有效应用于多种网络场景.因此,针对不同的网络应用场景,出现了多种SDN转发平面抽象,且具有其定制的转发平面实现模型.然而,异构的实现模型具有不同的架构和实现特点,为SDN网络研究带来极大挑战.提出一种可重构的SDN转发平面实现模型REFINE(Reconfigurable Forwarding Plane for SDN),通过提高灵活性从而支撑多样化SDN转发平面的实现.与传统的实现模型相比,它灵活性高、实现快速且成本低.通过分析目前SDN转发平面的处理流程将SDN转发平面实现模型分为八个功能模块,并根据这些模块的具体实现与转发平面抽象的相关性将REFINE模型分为抽象相关逻辑和抽象无关逻辑.通过重构抽象相关逻辑并使用统一接口可以灵活快速低成本地实现不同的SDN转发平面.原型系统表明REFINE可以灵活有效支持多种SDN转发平面抽象的实现.
[Abstract]:SDN Software-Defined networking) forwarding plane abstraction is the key to improve network control. Because of the limited ability of Open flow description, it is difficult to be applied to many network scenarios. Therefore, for different network application scenarios, there are many kinds of SDN forwarding plane abstraction, and it has its customized forwarding plane implementation model. However, the heterogeneous implementation model has different architecture and implementation characteristics, which brings great challenges to the research of SDN networks. A reconfigurable forwarding plane implementation model (REFINE) for reconfigurable forwarding plane (SDN) is proposed, which supports the implementation of multiple SDNs by improving flexibility. Compared with the traditional implementation model, it is flexible, fast and low cost. By analyzing the current processing flow of the SDN forwarding plane, the realization model of the SDN forwarding plane is divided into eight functional modules. The REFINE model is divided into abstract correlation logic and abstract independent logic according to the correlation between the implementation of these modules and the abstraction of forwarding plane. By reconstructing the abstract correlation logic and using the unified interface, different SDN forwarding planes can be implemented flexibly, quickly and cheaply. The prototype system shows that REFINE can support the implementation of multiple SDN forwarding plane abstractions flexibly and effectively.
【作者单位】：国防科学技术大学计算机学院;
【基金】：国家自然科学基金项目(61202483)资助国家“九七三”重点基础研究发展计划项目(2012CB315906)资助国家“八六三”高技术研究发展计划项目基金项目(2011AA01A10)资助
【分类号】：TP393.02

【参考文献】