Dynamic Query Optimization on a Distributed Object Management Platform

Fatma Ozcan


A Distributed Object Management (DOM) architecture, when used as the infrastructure of a multidatabase system, not only enables easy and flexible interoperation of DBMSs, but also facilitates interoperation of the multidatabase system with other repositories that do not have DBMS capabilities. This is an important advantage, since most of data still resides on repositories that do not have DBMS capabilities.

In this paper, we describe a dynamic query optimization technique for a multidatabase system, namely MIND, implemented on a DOM environment. Dynamic query optimization, which schedules intersite operations at run-time, fits better to such an environment since it benefits from location transparency provided by the DOM framework. In this way, the dynamic changes in the configuration of system resources such as a relocated DBMS or a new mirror to an existing DBMS, do not affect the optimized query execution in the system. Furthermore, the uncertainty in estimating the appearance times (i.e., the execution time of the global subquery at a local DBMS) of partial results are avoided because there is no need for the dynamic optimizer to know the logical cost parameters of the underlying local DBMS.

In scheduling the intersite operations a statistical decision mechanism is used. We also consider the schema integration information to make room for further query optimization. For this purpose, a method is presented that forms global query graphs by taking the schema integration information into account which is then used by the dynamic query optimizer. The proposed scheme tries to exploit the inherent parallelism in the system as much as possible.

The performance of the developed method is compared with two other most related techniques and the results of the experiments indicate that the dynamic query optimization technique presented in this paper has a better performance.

Joint work with S. Nural, P. Koksal, C. Evrendilek, A. Dogac.

Appeared in Proc. of Intl. Conf. on Knowledge and Information Management, Maryland USA, November 1996, pp 117-124.

Available in postscript.

Back to the Fall 1997 dbchat index.