Research And Implementation Of Key Techniques For Query Rewriting In Column-Store Data Warehouse

With the good effects of the wide application of the data warehouse, Data warehouse technology has developed rapidly. As the most important and commonly statement in data warehouse, Query execution efficiency directly affects the performance of data warehouse. In the query processing, Query is rewritten in another efficient internal representation, the query optimizer gain a better execution plan has been the academic research focus. For the frequent analysis and read features of the data warehouse, it is well known that query only needs to read the related column data in column-store system, avoiding the other columns data that is not related to query is read into memory, that column-store architecture is more suitable to data warehouse. It is necessary to explore a efficient query rewriting technology in column-store system.In this paper, firstⅠdescribe the research background of this issue and related technologies, highlighting the necessity of in-depth study to query rewriting in column-store system.Second,Ⅰintroduce some related important points of the issue, including data warehouse, the advantages of column-store, DWMS(column-store data warehouse system), Query processing. Traditional optimization strategies, etc.Third,Ⅰanalyse the characteristics of column-store system, and compare the difference of the query optimization between the column-store and row-store, propose some column-specific basic optimization strategies and rewrite rules. In the processing of group by operation, sub-query, the index connect, I do an in-depth analysis and discussion and propose a suitable solution, this further improved the query rewriting key techniques in column-store.Then, I present involved basic data structure in implementation, the rule-based query rewriting module is divided into three parts:rewriting the ordinary where query tree, rewriting the query tree with group by, rewriting the query tree with sub-query. And I propose some important implementation ways in detail.Finally, I put out column-store data warehouse management system DWMS3.0 as a platform, use SSB data sets, demonstrate key technologies mentioned in this article. The experimental results show that the proposed technologies can effectively reduce the size of the candidate plan, and exclude a large number of plans which can not generate the optimal plan, so as to make the cost and implementation time of query processing greatly reduced.