Data warehouse stream view update with hash filter.

A data warehouse usually contains large amounts of information representing an integration of base data from one or more external data sources over a long period of time to provide fast-query response time. It stores materialized views which provide aggregation (SUM, MIX, MIN, COUNT and AVG) on some measure attributes of interest for data warehouse users. The process of updating materialized views in response to the modification of the base data is called materialized view maintenance. Some data warehouse application domains, like stock markets, credit cards, automated banking and web log domains depend on data sources updated as continuous streams of data. In particular, electronic stock trading markets such as the NASDAQ, generate large volumes of data, in bursts that are up to 4,200 messages per second. This thesis proposes a new view maintenance algorithm (StreamVup), which improves on semi join methods by using hash filters. The new algorithm first, reduce the amount of bytes transported through the network for streams tuples, and secondly reduces the cost of join operations during view update by eliminating the recompution of view updates caused by newly arriving duplicate tuples. (Abstract shortened by UMI.)Dept. of Computer Science. Paper copy at Leddy Library: Theses & Major Papers - Basement, West Bldg. / Call Number: Thesis2003 .I85. Source: Masters Abstracts International, Volume: 42-05, page: 1753. Adviser: C. I. Ezeife. Thesis (M.Sc.)--University of Windsor (Canada), 2003