next up previous contents index
Next: Outline Up: T. Zurek: Optimisation of Previous: Histograms and IP-Tables

Performance Model

  



After having discussed the initial stage of the optimisation process in chapter 7, we now want to move on to stage 3 in which the performance of a partition is determined. The reason for skipping stage 2 at this point is that we require a good and thorough understanding of the cost implications for designing partitioning strategies that are to be employed in stage 2. In order to create efficient partitions we need to know what the expensive parts of partitioned temporal join processing are and how these are influenced by the choice of a partition.

The purpose of this chapter is to create a performance model for partitioned temporal join processing. This model will not only enable us to design efficient partitioning strategies but it is also an integral part of the optimisation process: it is the optimiser's principle tool for deciding on the quality of a particular partition.



 


Thomas Zurek