摘要
随着海量异构的数据不断进入数据仓库和系统用户的大量增加,在数据质量、可用性和查询等方面的因素将严重影响数据仓库的性能,所以数据仓库必须设计成可扩展的体系结构。文中采用可扩展的软件并行和硬件并行相结合的方法进行数据仓库的性能扩展,在数据仓库初建时采用SMP结构,当数据仓库增长到一定的时候采用高速缓存相关的非一致性内存访问结构,并且较好地利用I/O并行性,取得较好效果。使得当大量异构数据涌入可扩展数据仓库中时系统性能不会下降,很好地满足决策支持。
Along with mass heterogeneous data and system users getting into data warehouse, its performance begins to decline on data quality, usability and query aspects, thus data warehouse should be designed with scalable architecture. Sealable software and hardware technologies are used to extend test data warehouse in this paper, and in the beginning SMP is used in data warehouse. CC - NUMA is employed for extending data warehouse when it increases to a certain scale that SMP can not satisfy system performance need, and at the same time I/O parallel is used fully. In this case, a good effect on system performance is obtained when a great deal of heterogeneous data swarming into data warehouse, and meets the needs of decision support better.
出处
《计算机技术与发展》
2009年第5期65-67,71,共4页
Computer Technology and Development
基金
安徽省自然科学基金项目(2006KJ066B)
