الفهرس 
AbstractOnly 14 pages are availabe for public view
 |  | from | 152 |  |  |
| | | from | 152 | | |
Abstract
Recently, there is a growing need for query optimization algorithm that can effectively deal with federated database systems. Modern optimizers use a cost model to choose the best query execution plan (QEP) which heavily dependent on statistics maintained in the system catalog. Keeping such statistics up to date in the federation is troublesome due to local autonomy. A major challenge for global query optimization in federated systems is the lack of runtime information about the QEP selected within the remote datasources within the global level due to local autonomy. A learning optimizer for federated systems (DB2 II) extends Learning optimizer (LEO) feedback loop towards the federated architecture. Since there is no runtime monitor component for remote datasources, several approaches had been developed to get this run time information, but these approaches violates federation local autonomy by prohibiting the remote datasource optimizer from choosing the optimal QEP which might be totally different from the federated optimizer QEP, Consequently, this may lead long query execution time. This thesis proposes query processing and optimization framework to provide the following capabilities: Federated optimizer can accurately estimate the cardinalities, Preserving local autonomy of remote datasources by permitting remote optimizers to select their own QEPs, Permitting mid-query reoptimization. The proposed framework will grantee the federated learning optimizer set of unique features. It consists of seven phases which are as follow: (1) Preparation phase, (2) Planning phase, (3) F.M. phase, (4) DRDPE phase, (5) Local execution phase, (6) Analysis phase and (7) UDI execution phase. The federation monitoring (F.M.) and determining remote datasource planning and execution phases (DRDPE) are central to LEO since they provide the necessary abstraction of the federated DBMS execution system in terms of runtime information, as well as an abstraction of the QEP. So the promised objectives stated previously can be achieved by suggesting an algorithm that will help in overcoming the limitations of DB2 II in the federation. An investigation of the feasibility of the proposed framework through simulated federated database systems of interconnected datasources is presented. Several queries that use different datasources are simulated to measure the performance of the proposed framework. Experimental results show that: Using the proposed framework as a platform for running global queries improves the query execution speed compared to DB2 II.