覆盖多播节点失效检测的分布式算法的改进

An Improved Distributed Algorithm for Node Failure Detection Mechanism of Overlay Multicast

覆盖多播在端系统间构建数据分发树.由于端系统随意地加入或退出多播组,其可靠性不如路由器.当分发树中的非叶节点退出或者节点失效时,其下游节点将受到影响.低代价而快速的节点失效检测能降低由于服务间断所带来的影响,同时也为后续的恢复工作提供一个良好的基础.在分析相关工作的基础上,提出了一种改进的节点失效检测机制,它能够兼顾节点失效检测时间和额外消息负载两个性能指标.

Overlay Multieast constructs a data distributing tree among end systems. End systems may leave the multieast group voluntarily, so they are more susceptible to failures than touters. The non-leaf nodes in the distributing tree leave or fail, all downstream nodes are affected. Fast detection with low overhead can reduce the influence of the disruption of service, and provide a favorable basis for recovery in succession. Based on the related work, we propose an improved failure detection mechanism, which can greatly reduce the failure detection time, at the cost of slightly increased overhead.