分布式并行服务器的动态重构容错算法  

Dynamic reconfiguration fault-tolerance algorithmin a distributed parallel server

作　　者：左朝树 刘心松 邱元杰 郝尧 朱相文 

机构地区：[1]电子科技大学8010研究室,四川成都610054 [2]北京理工大学机电工程学院

出　　处：《系统工程与电子技术》2005年第5期910-913,共4页Systems Engineering and Electronics

基　　金：四川省科技攻关项目基金资助课题(02GG006-018)

摘　　要：分布式并行服务器的效率和可用性日益受到人们的关注。鉴于此,从体系结构上提出了动态重构容错模型,阐述了容错模型的数据分布和操作,并在此基础上描述了动态重构容错算法。该算法通过动态重构数据分布和操作解决了系统节点和网络故障,且使正在执行的任务不被中断。最后对比性地分析和测试了该算法的平均响应时间,说明该算法性能明显优于类似算法。此算法已成功应用于电子政务系统平台———分布式并行数据库系统DPSQL,经使用表明该算法使系统可用性和效率得到大幅度提高。Efficiency and availability in the distributed parallel server are being emphasized increasingly. In view of this, a dynamic reconfiguration fault-tolerance model is presented in view of system architecture, and the data distribution and operations in the model are described. On the basis of the model, a dynamic reconfiguration fault-tolerance algorithm is put forward and depicted in detail. The algorithm copes with the node and network faults by data distribution and operations of dynamic reconfiguration, meanwhile the tasks being executed will not be interrupted. Finally, the algorithm's average response time is comparatively analyzed and tested, which shows that the algorithm's performance is better than that of other similar algorithms. The algorithm has been successfully applied to the system platform of e-government affair - distributed parallel database system (DPSQL). The application proves the algorithm can greatly enhance the efficiency and availability of the system.

关 键 词：分布式并行服务器 动态重构 容错 可用性 

分 类 号：TP302.8[自动化与计算机技术—计算机系统结构]