Query optimization with value path materialization in column-stored DWMS

Abstract

Materialization is a key issue for query execution in column-oriented Data Warehouse Management System (DWMS) due to the fact that it has direct influence on the query efficiency. Focusing on the defects of traditional strategies Early Materialization and Late Materialization, this paper propose a new materialization strategy called VPMS (Value path Materialization Strategy) to solve those problem. First, VPMS define a new descriptor structure---pass block for the intermediate results during physical execution. For a given physical query tree, VPMS generates value path. Depending on the value path, the values of the column are saved in the value area of the pass block when needed by the upper nodes, otherwise, only save the location information. Finally, during the query execution, the physical sub-operation is specified according to the materialization path and the query object itself, which effectively reduces the unnecessary duplication of judgments. Experiments on benchmark dataset SSB demonstrate the overall effectiveness of our approach.