Adaptive query evaluation: new challenges from internet-based applications
- Speaker: Dr Fabio Porto (Ecole Polytechnique Federale de Lausanne, Switzerland)
- Host: Alvaro A. A. Fernandes
- 19th October 2005 at 14:15 in 1.5
Database query processing approach has evolved to offer traditional applications a secure and efficient environment for storing and retrieving data. A major aspect of this approach is the transformation of complex SQL queries into a tree of simple operations, which is usually done in a pre-processing stage based on cost estimates. A cost model is supposed to predict run-time conditions and indicate a quasi-optimal query execution plan. This talk will explore new challenges in query processing that arise when assumptions about future behavior are violated. Reasons for this are manifold. Some scientific applications run over data and programs available in the internet and for which few or no metadata exists. The adoption of a grid environment for the evaluation of long running queries inherently assumes resource heterogeneity and variability on their availability. In this context, dynamic monitoring is a necessity and decisions must be taken to adapt query execution to run-time conditions. In the second part of this talk, I will present CoDIMS-G which is an adaptive query processing service developed to support scientific applications in the grid environment and discuss the main adaptation issues we have addressed in the system.