一种数据库中间件的配置服务的研究和实现  

Implementation of Configuration Service of a Database Middleware

作　　者：宋怀明[1] 焦丽梅[1] 孙凝晖[1] 

机构地区：[1]中科院计算所国家智能计算机中心

出　　处：《小型微型计算机系统》2007年第3期438-442,共5页Journal of Chinese Computer Systems

基　　金：北京科技计划项目(H030130020330)资助.

摘　　要：在基于事件流的大规模数据密集型系统中,数据可分为事件流数据和事件配置数据两大类,配置数据表示事件流的规则.在shared-nothing结构下,配置数据一般采用全复制的方式分布到各个数据库节点,用于和事件流数据的联合查询.采用全复制的配置数据,修改操作必须在所有节点上进行,数据的一致性控制和多节点的事务处理成为此类数据管理的关键问题.对配置数据的特点及其管理策略进行分析,并成功的在DBroker系统中实现了配置数据一致性控制.In a streaming llke large-scale data-density system, data can be classified as two major kinds： one is called event streaming, and the other is called event configuration, which is recorded as rule of event streaming data. In a shared-nothing architecture, configuration data is replicated in all database nodes, each node with a full copy. So any modifying of configuration data must be processing in all nodes, thus data consistency control and transaction processing becomes an extremely impor- tant issue. This paper proposes a detailed analysis of replicated data consistency control and transaction processes in large-scale systems based on shared-nothing architecture, as well as proposes the strategies applied in a practical Large-scale Data-density System named DBroker.

关 键 词：shared—nothing DBroker 大规模数据密集型系统 数据库中间件 多复制一致性 分布式事务 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]