摘要
状态检测是高可用性和可靠性的重要组成部分 ,设计实现了一个分布式的高可用检测机制 ,该机制逻辑上分为一个主节点和多个从节点 ;主节点由所有节点选举产生 ,周期性的对需要检测的虚拟子网进行轮循 ,从节点在收到轮循消息后给予响应 ;主从节点分别依据轮循响应和轮循信息对彼此的健康状态进行检测 .较好的定时策略 ,确保了较低的误报概率和诊断延迟 .该机制支持负载在主从节点上的均衡分配 ,扩展性好 ,状态机需要维护的状态数少 ,管理简单 .
A detection mechanism of the nodes or daemon failures in the network was presented in this paper as the key technology for high availability of a clu s ter network. It composes of a master node and several standby nodes. The master node is elected by all nodes which run the same process but in different states. Master sends a message (Polling) to every other Standby process periodically an d then waits to receive a reply message (Polling ACK) from every other process. Master node and standby nodes can analyze the health state with each other by Po lling ACK and Polling messages respectively. Timer policy enables lower probabil ity of premature termination, less detection delay. The mechanism supports also load balancing and nice extensibility.
出处
《小型微型计算机系统》
CSCD
北大核心
2005年第1期11-14,共4页
Journal of Chinese Computer Systems
基金
国家"八六三"重点项目
新型网络服务器系统 (2 0 0 1AA11112 0)
关键词
容错
任务迁移
可用性
负载平衡
可扩展性
心跳
fault tolerance
fail-over
availability
load balancing
extensibility
heartb eat
