摘要
Query efficiency is bottleneck of XML data cube aggregate query. pXCube is a kind of XML data cube model based on path calculation. Join operations are avoided in this model, but the query efficiency of fact cell is become a new bottleneck. This paper focuses on parallel technology of cloud computing to improve query efficiency of pXCube. Mixed partitioning strategy for fact and dimensions is applied in pXCube cloud model, and the same partitioned vector is adopted. Query parallel algorithm of pXCube cloud model is presented as well. Experiments show that the query cost of pXCube cloud model decreases with the increasing number of parallel nodes gradually. The query cost of fact fragments of each node are close to or even lower than join operations of dimensions, and the Speedup is with better linear. So the model is well suited for decision supported query.
Query efficiency is bottleneck of XML data cube aggregate query. pXCube is a kind of XML data cube model based on path calculation. Join operations are avoided in this model, but the query efficiency of fact cell is become a new bottleneck. This paper focuses on parallel technology of cloud computing to improve query efficiency of pXCube. Mixed partitioning strategy for fact and dimensions is applied in pXCube cloud model, and the same partitioned vector is adopted. Query parallel algorithm of pXCube cloud model is presented as well. Experiments show that the query cost of pXCube cloud model decreases with the increasing number of parallel nodes gradually. The query cost of fact fragments of each node are close to or even lower than join operations of dimensions, and the Speedup is with better linear. So the model is well suited for decision supported query.
基金
supported by National Natural Science Foundation of China under Grant No. 61072091
