الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 16 |  |  |
| | | from | 16 | | |
Abstract
Applying workflow technology to data warehouses can enhance multitude of its functions. Workflow technology can streamline and coordinate the different data warehouse processes and help in defining and controlling situations like handling of failures and adding new sources of information. In this thesis, a modified workflow model based on the DOM (Distributed Object Management) model is proposed. The proposed model adds new control flow constructs to the DOM model that are necessary to specify a workflow. The full syntax, graphical notation and semantics of the modified DOM workflow is presented. A distributed scheduling mechanism is suggested to control execution of the workflow by the use of guards to enforce dependencies. A mapping technique between the suggested specification and guards is suggested. A failure handling technique is introduced. It is a local forward recovery technique that resubmits the aborted subtask in case of task failure.
The key features of the model are; introducing simple and easy way to specify workflows, providing distributed scheduling which is necessary to meet the needs of various current workflows, and the scheduling process is made simple by the suggested mapping between ACTA constructs and guards.
To test the proposed model, it is applied to the Sweep view maintenance process in data warehouses. A full specification of the Sweep using the proposed model is presented along with the generation of needed guards. A comparison between representing Sweep by both the original DOM model and the suggested model is introduced. A simulation model for the Sweep process using the proposed workflow is proposed. The model allows representing the process of incrementally updating the data warehouse while serving its normal query load at the warehouse, i.e. the model resembles a real warehouse system. This work is new since even without a workflow, the performance, scalability, and recoverability of incrementally updating the warehouse had not been investigated by the research community.