PiBuffer:面向数据中心的OpenFlow流缓存管理模型

发布时间：2018-06-08 03:44

本文选题：数据中心网络 + 软件定义网络　；参考：《计算机学报》2016年06期

【摘要】：基于OpenFlow的SDN(Software Defined Networking)技术在数据中心中得到广泛研究和应用,如何缓解集中的控制平面成为网络性能的瓶颈是其中的研究热点.OpenFlow规范提出,当数据平面有缓存能力时,未命中的报文仅需发送少量摘要信息至控制器触发规则下发,从而减少控制平面与数据平面的通信负载.然而,现有的缓存模型采用报文粒度的缓存方式,使得同一条流的多个未命中报文会被送至控制器造成额外的通信负载,而且交换机处理报文的顺序会导致流内报文乱序,从而降低通信的性能.针对上述问题,该文提出了一种支持流内报文保序的OpenFlow交换机流缓存管理模型.通过基于流粒度的未命中报文缓存方式,进一步减少控制平面与数据平面的通信开销.通过设计流动作预处理机制,实现同一条流内报文传输保序.该文分别基于软件交换机OFSoftSwitch与硬件网络实验平台NetMagic对该流缓存管理模型进行了原型系统验证.
[Abstract]:SDN Software defined networking (SDN) technology based on OpenFlow has been widely studied and applied in the data center. How to alleviate the centralized control plane becomes the bottleneck of network performance is one of the research hotspots. OpenFlow specification proposes that when the data plane has the ability to cache, The missed message only needs to send a small amount of summary information to the controller trigger rule to reduce the communication load between the control plane and the data plane. However, the existing cache model adopts packet granularity caching method, which causes multiple missed packets of the same stream to be sent to the controller to create additional communication load, and the order in which the packets are processed by the switch will cause the packet order in the stream to be chaotic. Thus, the communication performance is reduced. In order to solve the above problems, this paper proposes a flow cache management model for OpenFlow switches. The communication overhead between the control plane and the data plane is further reduced by using stream granularity based missed packet cache. By designing the preprocessing mechanism of the stream action, the message transmission in the same stream can be carried out in order. Based on OFSoftSwitch and NetMagic, the model of stream cache management is validated in this paper.
【作者单位】：国防科学技术大学计算机学院;
【基金】：国家“八六三”高技术研究发展计划项目基金(2015AA016103) 国家自然科学基金(61202483)资助
【分类号】：TP393.07

【相似文献】