Validated Cost Models for Parallel OQL Query Processing (2002)

Author(s): de F Mendes Sampaio S, Paton NW, Smith J, Watson P

    Abstract: Query cost models are widely used, both for performance analysis and for comparing execution plans during query optimisation. In essence, a cost modelp redicts where time is being spent during query evaluation. Although many cost models have been proposed, for serial, parallel and distributed database systems, surprisingly few of these have been validated against real systems. This paper presents cost models for the parallel evaluation of ODMG OQL queries, which have been compared with experimental results obtained using the Polar object database system. The paper describes the validation of the cost model for a collection of queries, using three join algorithms over the OO7 benchmark database. The results show that the cost model generally both ranks alternative plans appropriately, and gives a useful indication of the response times that can be expected from a plan. The paper also illustrates the application of the cost model by highlighting the contributions of different features and operations to query response times.

      • Date: 2-5 September 2002
      • Conference Name: 8th International Conference on Object-Oriented Information Systems (OOIS)
      • Volume: 2425
      • Pages: 60-75
      • Publisher: Springer
      • Publication type: Conference Proceedings (inc. abstract)
      • Bibliographic status: Published

      Professor Paul Watson
      Director of the Digital Institute, Professor of Computer Science