الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 16 |  |  |
| | | from | 16 | | |
Abstract
A data warehouse collects and maintains a large amount of data from mull pIe distributed autonomous information sources, in fonnS O views, in order to provide fist access to the integrated data. One of the major problems in data warehousing is materialized view maintenance. Although much attention has been given to this problem, existing solutions assumed that both information sources and the data warehouse are stationary. introducing mobility n a data warehousing systems poses its unique challenges that have not been considered in traditional view maintenance algorithms. This thesis introduces two novel algorithms for incrementally maintaining materialized views defined over information sources and stored in a data warehouse, where the information sources and/or the data warehouse can be mobile. n the proposed algorithms, updates are pushed in batches and propagated along information sources until the view change is finally calculated and delivered to the data warehouse. in contrast to other existing approaches, the proposed algorithms have the advantage that the workload of updating the view is distributed among in{brmation sources alleviating the data warehouse from receiving and handling unnecessary intermediate messages befbre the view change is finally calculated. The proposed algorithms ensure strong consistency of the view at the data warehouse in the presence of concurrent updates, while reducing the number of messages needed, the communication overhead, and the time needed for the view updating process. Moreover, the FIFO assumption in pervious algorithms is relaxed. Further, they do not require the sources or the data warehouses to be connected during the whole update process, which is suitable for mobile environments. The performance of the proposed algorithms is evaluated through the implementation of a simulation model adapted from the WHIPS architecture of Stanford University.