Font Size: a A A

The On-line Consistency Of Materialized View Maintenance In A Warehousing Environment

Posted on:2003-09-16Degree:DoctorType:Dissertation
Country:ChinaCandidate:J Y ChenFull Text:PDF
GTID:1118360092465720Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
The Data Warehouse movement is a consequence of the competitive business pressures. And the data warehousing, targeted for decision support, is a repository of subject-oriented data that has been cleaned up, extracted and transformed from many source in transaction processing. Since it was firstly introduced in the nineties of last century, the data warehouse movement has been become the swim in many industries. Moreover, the data warehousing technologies have been become the second hotshot technique following the Internet technique in American.As one of the hotshot techniques, materialized views on-line maintenance is an emerging technique of views maintenance in a warehouse. That is, when source data changes, warehouse views, which are in using simultaneity, need to be maintained promptly and correctly so that they remain consistent with the source data.Commercial data warehousing systems typically refresh all warehouse views periodically to keep them up to date. In the other words, the system compiles the updating records, and refreshes the warehouse views periodically to keep them up to date, while the warehouse was enjoined. So in order to guarantee the end users to read the data of warehouse in working time, many commercial systems refresh the date warehouse at night. But there are three questions as follows:Firstly, many customers have international operations in multiple time zones, so there is no convenient down time, no "night" or "weekend" when new sources or views can be added and all of the recent updates can be batched and processed together to (re) compute materialized views.Secondly, because of the large volumes of data in update, the more time is spent in maintenance, but the warehouse loads must be completed within a fixed period, while the warehouse is not available. So enough time must be offered for a new load.Thirdly, the strategy of periodical maintenance can make the data out of date. Specially, for some assignments, such as war field, the off-line periodical loads can't be accepted in this data warehouse maintenance.In one words, 24×7 Operation in warehouse maintenance is an interesting problem in the research of the materialized views on-line maintenance.For the materialized views on-line maintenance in a single source model, a extended schema of source relation is intruded to label the updated records of source;And using version control and compensating technique, along with acknowledgement mechanism, the authors introduce a new algorithm, Glide, to synchronize the maintenance process between the database and the data warehouse so as to ensure the data consistency; The algorithm Glide is divided into Glide-DW and Glide-DB, which execute in database and warehouse separately; And the algorithm ensure the data consistency under the invention of OLAP query; Several results are improved, and the robustness of the algorithm or the utilization ratio of CPU in source computers achieves a high performance; In addition, a proof is given that the level of consistency in algorithm Glide is complete; And the author illustrates the application of the algorithm by a typical example; If a view includes a key from every base relation involved in the view, then the author streamline the ECA-Key algorithm to solve this problem.Maintaining the consistency of warehouse data is challenging, especially if the data sources are autonomous and views of the data at the warehouse span multiple sources, since the views may not be obvious when new updates from a source impact the processing of previous updates. If a view includes a key from every base relation in the view, the algorithm Strobe and the improvement algorithm of Strobe, Strobe*, are introduced in the paper, and the proof of Strobe* is given, which is strongly consistent; If a view is in general, the author gives a wide algorithm, M-Glide, which is strongly consistent, to solve this problem, and the algorithm is a essential extend of the algorithm Glide; The algorithm M-Glide imposes an action list to record the updating...
Keywords/Search Tags:view maintenance, warehouse, consistence, algorithm
PDF Full Text Request
Related items