摘要
通过分析数据中心网络特征和云计算应用需求,提出一种适合数据中心网络的可靠组播机制R2M。针对云计算应用同时强调数据完整性和传输实时性的要求,R2M充分利用数据中心网络服务器间存在多路径的特点,通过组织多棵不相交组播树并行传输数据,并在相邻节点间采用P2P技术建立跨树的恢复关系,实现恢复路径与组播传输路径完全隔离。选择最邻近节点重传数据和较小的路径差错相关性使得R2M能够快速恢复丢失的数据包。基于BCube的仿真实验结果显示R2M具有高可靠性和较小的恢复延迟。
A novel reliable multicast approach named R2M for data center network is proposed by analyzing the characteristics of data center network and requirement of cloud computing applications. By leveraging multipath in data center network, R2M constructs multiple diverse multicast trees to transmit packets simultaneously and builds isolated repair path between nearby group members utilizing P2P technology for packet recovery. As a result, R2M improves error resilient ability in the presence of various levels of packet loss caused by unpredictable congestion, node failure and tree-reconstruction. The results of simulations indicate that R2M achieves low latency and high reliability.
出处
《计算机工程与设计》
CSCD
北大核心
2012年第8期2966-2970,共5页
Computer Engineering and Design
基金
北京工业大学博士科研启动基金项目(00700054R1760)
关键词
数据中心网络
可靠组播
多路径
云计算
快速恢复
data center network
reliable multicast
multipath
cloud computing
rapid recovery