一种维护分解的并行增量视图维护方法  

Parallel Method for Incremental View Maintenance Based on Decomposition of Maintenance Task

作　　者：左亚尧[1,2] 舒忠梅[1] 汤庸[1,3] 

机构地区：[1]中山大学信息科学与技术学院,广东广州510275 [2]广东工业大学计算机学院,广东广州510090 [3]华南师范大学计算机学院,广东广州510631

出　　处：《小型微型计算机系统》2011年第2期279-285,共7页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(6037308160673135)资助;广东省自然科学基金重点项目(4105503)资助;广东省自然科学基金项目(06300527)资助

摘　　要：实化视图维护是指在数据源的原始数据发生改变时,有效地将这种变化反映到数据仓库中,使相应的实化视图得到及时更新.当前的视图维护方法主要以C/S结构为基础,当更新频繁时将会导致数据仓库超载而崩溃.针对这种现状,提出基于维护查询任务进行分解的实化视图并行增量维护P3Sweep算法,算法修正了Sweep算法只能顺序处理的限制,赋予其并行处理能力.对于单个更新,P3Sweep算法通过对维护查询任务进行分解,并行执行左右扫描过程来完成维护演算;对于并发更新,算法通过递归分解方式进行细粒度并行演算.性能计算和实验测试结果表明,该方法增加了系统对更新的吞吐能力,减少了维护的延迟,从维护查询任务本身实施对实化视图维护的优化.View maintenance is an important research topic in data integration technology. Currently, most view maintenance methods are based on the C/S model, which will lead to the overload and crash at the data warehouse. After analyzing the current online incremental view maintenance methods, based on the decomposition of maintenance query task, a parallel incremental maintenance algorithm called P^3Sweep is proposed. The algorithm optimizes the performance of the Sweep algorithm from two aspects. On the one hand, p^3 Sweep executes the left and fight sweeps in parallel for one single update. On the other hand, p3 Sweep recursively incorpo- rates all the concurrent updates encountered during the evaluation of an update. The algorithm recursively evaluates a relevant concurrent update by suspending the current evaluation. According to performance evaluation and the experiment results, the algorithm maintains view consistency with the least time overhead and promotes the efficiency greatly.

关 键 词：实化视图 维护查询 分解 并行 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]