The Fitness Benchmark for Data Warehousing

Abstract

This paper challenges the notion that speed of query execution, which dominates the OLTP database benchmarking landscape, is the appropriate critical metric for benchmarking data warehouses. The heart of the work is the introduction of a new fitness benchmark for testing and rating a warehouse’s ability to provide decision support rather than testing its ability to quickly respond to queries. The proposed benchmark contains three specific metrics: utilization (thorough use of existing data structures), coverage (non-reliance on OLTP source systems) and granularity (robust support of required aggregations). The metrics were chosen for their desirability, impactfulness, portability, simplicity and ease of improvement. The paper defines the metrics in detail, identifies the metadata used in evaluating each of the measures, proposes and defines a weighting scheme for combining them into the overall benchmark and demonstrates the computation of the benchmark against a well-documented, publicly available reference set of decision support data. The paper concludes that using the results of the computed benchmark to direct specific change can lead to measurable improvements in the ability of a warehouse to provide decision support.

13 Figures and Tables

Cite this paper

@inproceedings{Goshey2006TheFB, title={The Fitness Benchmark for Data Warehousing}, author={Michael Goshey}, year={2006} }