基于流粒度的OpenFlow分组缓存管理模型

发布时间：2018-06-07 19:43

本文选题：数据中心网络 + 软件定义网络　；参考：《计算机工程》2017年02期

【摘要】：基于OpenFlow的软件定义网络(SDN)技术通过在OpenFlow交换机中建立有效的缓存模型,能够大幅减少控制平面和数据平面的通信负载,但整条数据流的缓存模型会对数据流的传输造成较大延时,降低整个SDN的数据传输性能。针对该问题,引入PiBuffer流缓存模型,构建基于报文分组粒度的分组缓存模型。通过在控制平面建立流路由和流状态的缓存信息,分别对流报文之间和交换机之间的数据传输采用"分组缓存,组内保序"机制和"传输询问,完成通知"机制,并对控制平面和数据平面的通信消息进行优化,以提高数据中心网络的通信性能。软件模拟结果表明,在数据中心基于OpenFlow技术的SDN网络中,该模型比流缓存模型具有更优的通信性能。
[Abstract]:Software definition Network (SDN) technology based on OpenFlow can greatly reduce the communication load between the control plane and the data plane by establishing an effective buffer model in the OpenFlow switch. However, the buffer model of the whole data stream can delay the transmission of the data stream and reduce the data transmission performance of the whole SDN. To solve this problem, PiBuffer stream caching model is introduced to construct packet caching model based on packet granularity. By establishing the stream routing and stream state cache information on the control plane, the data transmission between packets and switches adopts the mechanism of "packet cache, group internal order preservation" and "transmission inquiry, complete notification" mechanism, respectively. The communication message between the control plane and the data plane is optimized to improve the communication performance of the data center network. The simulation results show that the model has better communication performance than the flow buffer model in the SDN network based on OpenFlow technology.
【作者单位】：上海理工大学光电信息与计算机工程学院;上海理工大学管理学院;
【基金】：国家自然科学基金(61572325,60970012) 高等学校博士学科点专项科研博导基金(20113120110008) 上海重点科技攻关项目(14511107902) 上海市工程中心建设项目(GCZX14014) 上海智能家居大规模物联共性技术工程中心项目(GCZX14014) 上海市一流学科建设项目(XTKX2012) 沪江基金研究基地专项(C14001)
【分类号】：TP393.0

【参考文献】