首页 | 本学科首页   官方微博 | 高级检索  
     

数据库集群的"无应答严次序发布传播"复制策略
引用本文:刘晖,沈钧毅,彭勤科,邢东山.数据库集群的"无应答严次序发布传播"复制策略[J].小型微型计算机系统,2003,24(12):2248-2254.
作者姓名:刘晖  沈钧毅  彭勤科  邢东山
作者单位:1. 西安交通大学,软件所,陕西,西安,710049
2. 西安交通大学,系统所,陕西,西安,710049
基金项目:NationalNaturalScienceFoundationofChinaWithGrant(No .60 1 750 1 5),JiaotongUniversity IBMButoneXi’anSoftwareCenter.
摘    要:数据库集群服务器的关键技术是数据复制,本文在对比分布式文件系统、可靠型多播、对象组服务以及分布式内存等技术的基础上,提出了“无应答严次序发布传播”复制策略,“发布传播”类似但不同于“发布订阅”模型。该策略采用了IP多播、双钥锁和基于全局/局部/状态次序编号的次序管理机制。双钥锁允许加锁者或下一个申请者释放互斥锁。次序管理遵循4个原则来避免“乱序”、“意图歪曲”、“因果牵制”和“反向因果牵制”四个数据不一致问题。本文详细讨论了该策略的设计目标、主要问题及解决途径、特征、基本原则以及算法轮廓。

关 键 词:集群  数据库集群服务器  数据一致性  互斥锁  执行次序  无应答严次序发布传播  复制策略

NASO Publish-propagate Replication Scheme for Clustered Database Server
LIU Hui ,SHEN Jun-yi ,PENG Qin-ke ,XING Dong-shang.NASO Publish-propagate Replication Scheme for Clustered Database Server[J].Mini-micro Systems,2003,24(12):2248-2254.
Authors:LIU Hui  SHEN Jun-yi  PENG Qin-ke  XING Dong-shang
Affiliation:LIU Hui 1,SHEN Jun-yi 1,PENG Qin-ke 2,XING Dong-shang 1 1
Abstract:Clustered Database Server (CDS) supports dynamic Web pages and allows online updating. Because each node of a CDS has the same copy of business-logical objects in spite of what atomic architecture is adopted, replication is essential. After commenting on some different techniques to achieve replication for CDS, such as DFS, reliable Multicast protocol, Groupware services and DSM, this paper puts forward a NASO Publish-Propagate Replication scheme. NASO stands for No-Acknowledgement-Strict-Order and Publish-Propagate is a paradigm close to but different from Publish-Subscribe paradigm. This scheme depends on IP Multicast protocol, Twin-Keys Locking mechanism and Sequence Management based on Global/Local/Status Order Numbers. By this replication scheme, either the lock-holder or the next applicant can unlock the corresponding exclusive locks. At the same time, four sequence-management rules are used to avoid divergence, intention violation, causal diversion and inverse causal diversion. This paper focuses on its design motives, main problems and their solutions, characteristics, basic rules and algorithm frames.
Keywords:cluster  database  replication  data consistency  locking  ordering
本文献已被 CNKI 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号