数据仓库环境中多级负荷的全局缓冲管理

GLOBAL CACHE MANAGER FOR MULTI-CLASS WORKLOADS IN DATA WAREHOUSING ENVIRONMENTS

数据仓库查询要求为数据库和导出数据提供有效的全局缓冲管理．本文描述一个兼顾数据库页和导出数据的全局缓冲管理器，提出收益量度用以比较缓存导出数据，为查询操作提供工作空间的效果，并在此基础上设计算法，为多查询生成有效的缓冲分配方案．分配算法与不同的替换策略结合形成一族全局缓冲管理方法，将其与简单LRU和LRU－K策略进行实验比较，结果表明本文提出的方法更适于数据仓库中的多级负荷．

Data warehousing queries require efficient global buffer management for both databases and derived data. This paper describes an efficient global cache manager that caches both database pages and derived data. A profit metric is developed to compare the expected benefit of caching derived sets and buffering for query operations. Then on the basis of this model, cache allocation algorithms are designed to produce efficient allocation scheme for multiple queries. These algorithms are combined with different replacement strategies to generate a family of approaches. The results of experiments to investigate their performances and to compare them with the simple LRU and LRU-K indicate that the approaches are adequate for multi-class workload in data wareh0using environments.