Based On Data Consistency Of The Data Warehouse Materialized View Maintenance

Usually, there are a large number of material views are stored in data warehouse, whose data is mainly used for decision analyzing. Because the data in the data warehouse are mostly extracted from operating databases, therefore, the data in the data warehouse should be updated regularly in order to keep accordance with the data in the data source. In maintaining material views, the quality of the algorithm will affect the efficiency of maintaining directly.The routine algorithms of maintaining views have exposed some efficiency questions in the consistency between material views and the lower data source layer, what is more, some algorithms will cause exception questions. Firstly, this paper regards the WHIPS systems, which was put forward by Stanford University, as foundation stone, and dissects its principle in detailed. Also, we have analyzed its system structure and function of each module. Secondly, this paper has introduced the routine algorithms systematically, and expounded the principle of each algorithm in detailed, we have pointed out their advantages and shortcomings, and improved each algorithm. Finally, this paper proposes a novel algorithm ~ Parallel View Maintenance (PVM) algorithm.PVM algorithm has synthesized several kinds of different algorithms, including Parallel Online Error-correcting (POLEC) algorithm, Dependence updating detect algorithm and Self view maintenance algorithm, This algorithm deals with view maintenance from different angles or aspects, and maintain the consistency of the views in the minimum time costs, So it can improve the maintenance capability greatly. When receiving updating command, PVM algorithm checks up it with dependence at first: As to the independent updating, PVM abandon it directly; To the dependent updating, it executes parallel increment view maintenance or self view maintenance mechanism checking; To the updating which meet self maintenance, it avoids network transmission andkeeps material views accordance with the lower data source by auxiliary views. To the updating that must be proceeded parallel increment view maintenance, it makes use of POLEC to handle it. To do so, this algorithm can increase the updating capability and decrease the delay of maintenance.In this paper, there are three mainly aspects of innovation as follows:Firstly, it revises the bottleneck of POLEC Algorithm in that which can only proceed view maintenance in proper order, and endow it the parallel processing ability, also this algorithm has make a deep analysis about problems of parallel processing and give a all-around solution project.Secondly, on the basis of POLEC, PVM algorithm incorporates the advantages of several different kinds of algorithms; it is the first time to introduce the thought of self-maintenance and relevance updating processing, to integrate these three kinds of different classifications algorithms as a whole.Finally, this algorithm inherits the advantages of all the past algorithms, and totally overcome the defects of them, thus it can get better response time and decrease the total amount of the messages. From making science estimates of PVM algorithm, we can assert that this algorithm has obvious promotion in view maintaining performance.