摘要
What-If分析能够提供比传统的OLAP(on-line analysis processing)分析更加有意义的决策支持信息.基于历史数据的应用场景假设分析需要更加有效的what-if数据视图生成机制的支持.在传统的delta表合并算法的基础上,提出了基于内存记录指针的deltaMap算法来提高what-if数据视图的合并性能.根据OLAP分析的应用特点,提出了pre-merge算法来处理支持分布式计算的聚集运算.根据不同的假设更新类型,对查询重写算法和△cube算法作了详细的性能测试并进行了全面的性能分析对比,在此基础上提出了what-if分析的代价模型,以应用场景模式、假设更新率、假设更新复杂度、查询结果集的基数作为参数,有效地描述系统what-if查询处理策略,为what-if分析的解决方案提供了一个可行的框架结构.
A what-if analysis can provide a more meaningful information than classical OLAP （on-line analysis processing）. Multi-Scenario hypothesis upon historical data needs efficient what-if data view support. Two novel algorithms of deltaMap and pre-merge, which can greatly improve the performance of delta table algorithm with set operations, are proposed. To analyze the performance of query re-writing algorithm and delta cube algorithm under different what-if update conditions, a global performance analysis and comparison are presented in the experiment section. This paper proposes a cost model for a what-if analysis processing engine, based on different algorithms with parameters such as application scenario, what-if update rate, complexity of what-if updates, memory storage policy, cardinality of query result set etc, that can be used as a practical framework in a what-if analysis system.
出处
《软件学报》
EI
CSCD
北大核心
2010年第10期2494-2512,共19页
Journal of Software
基金
国家自然科学基金Nos.60473069
60496325
国家高技术研究发展计划(863)No.2008AA01Z120
国际合作(HP实验室)项目
北京市教委产学研合作项目
中国人民大学研究生科学研究基金No.08XNG040~~
