软件定义Fat-Tree数据中心网络的多组播树动态切换机制

发布时间：2018-05-01 06:08

本文选题：多树组播 + Fat-Tree数据中心网络　；参考：《小型微型计算机系统》2017年04期

【摘要】：数据中心网络中单组播树传输往往难以应付网络拥塞和故障,从而导致可靠性和网络资源利用率较低.多组播树动态切换机制虽然能够灵活地对数据传输进行均衡和故障恢复,但在传统网络中难以部署和实现.软件定义网络将控制面和数据面分离,具有良好的可控性和可编程性.在支持Open Flow的Fat-Tree数据中心网络中,提出一种基于自适应滤波预测的多组播树动态分级切换算法.该算法首先为每个组播会话建立多个备选组播树,并根据收集的网络链路状态为每个组播树计算优先级值;然后采用自适应滤波算法对优先级值进行预测,并利用预测的优先级值设计组播树的动态分级切换策略和数据分发比例.最后,通过在Mininet平台进行了仿真实验,验证了本文所提算法的可行性和性能.
[Abstract]:In data center networks, single multicast tree transmission is often difficult to cope with network congestion and failure, resulting in low reliability and network resource utilization. Although the multi-multicast tree dynamic handoff mechanism can flexibly equalize and recover the data transmission, it is difficult to deploy and implement in the traditional network. The software defined network separates the control surface from the data surface and has good controllability and programmability. A multi-multicast tree dynamic hierarchical handoff algorithm based on adaptive filtering prediction is proposed in the Fat-Tree data center network supporting Open Flow. The algorithm firstly establishes several alternative multicast trees for each multicast session, and calculates priority values for each multicast tree according to the collected network link state, and then uses adaptive filtering algorithm to predict the priority values. The dynamic hierarchical handoff strategy and data distribution ratio of multicast tree are designed by using the predicted priority value. Finally, the feasibility and performance of the proposed algorithm are verified by simulation experiments on Mininet platform.
【作者单位】：中国科学技术大学自动化系;
【基金】：国家自然科学基金面上项目(61573329)资助中央高校基本科研业务费专项资金项目(WK2100100026)资助博士后基金项目(BH2100000005)资助
【分类号】：TP393.03;TP308

【相似文献】