基于马尔可夫预测模型的数据仓库缓存管理

Markov model prediction based cache management policy

提出了一种基于马尔可夫预测模型的数据仓库缓存管理策略.将Chunk作为缓存的基本粒度单位,通过收集用户已提交的查询,利用马尔可夫模型预测下一步用户将要访问的视图区域的概率分布,并在此概率分布的基础上提出了基于预测风险的缓存淘汰算法.实验结果表明,算法缓存命中率高,有效地缩短了OLAP查询的响应时间,提升了系统的整体性能.

In this paper,through employing the chunk as the essential granularity unit of cache,a new data warehouse cache management policy based on Markov model prediction was proposed.According to the query that user has submitted and Markov model to predict the query view probability distribution that will happen the next unit.Advancing a cache replacement algorithm RPB,basing on this probability distribution,which improve the utilization of the caching data in order to optimize the system performance.