基于集群的轻量可扩展TCP容错方法  

Cluster computing-based fault tolerance scheme of lightweight and scalable TCP connection

作　　者：邵志远[1] 金海[1] 

机构地区：[1]华中科技大学计算机科学与技术学院,湖北武汉430074

出　　处：《华中科技大学学报（自然科学版）》2006年第8期36-39,共4页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家高技术研究发展计划资助项目(2002AA1Z2102)

摘　　要：针对现有TCP容错方案在高可用性方面的不足,提出了一个新的容错方案———轻量拷贝的传输控制协议ER-TCP.该方案采用日志管理配合主动复制的方法,可降低通信转发结点的负载;采用高效的同步算法可以减轻多个备份服务器的同步对通信造成的性能损失.与其他TCP容错方案进行性能测试的结果比较表明,ER-TCP对通信造成的性能损失小,在多个备份服务器的情况下ER-TCP仍可保持较高的通信效率.There being poorer high availability in existing transmission control protocol （TCP）, a novel TCP fault-tolerance scheme, ER-TCP, is proposed. The load of the primary server is alleviated by the combination of logging mechanism with active replication, and the performance penalty from multiple backup servers synchronized is reduced by using a high efficient algorithm. The experimental resuits show that ER-TCP scheme exhibits its better performance than that of other schemes, keeping high communication efficiency with multiple backup servers being used.

关 键 词：集群计算 可靠性 高可用性 传输控制协议(TCP) 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]