Baron Schwartz, our CEO and founder, is speaking at the Performance and Capacity Conference by CMG on November 6th. CMG, the Computer Measurement Group is a not-for-profit organization that focuses on the efficiency and scalability of IT service delivery through measurement, quantitative analysis and forecasting. It has been a reputable, worldwide resource for best practices in the IT profession since 1974.
Baron will be contributing to these efforts when he presents on “Using Regressions to Estimate Query Resource Consumption.” VividCortex uses regression to generate metrics for many things that cannot be measured directly, allowing us to provide robust sets of data while reducing the measurement intrusion effect on systems. In particular, our customers are able to see the estimated CPU usage of a group of queries and better plan for capacity.
If you will not be attending the conference but are interested in the method and application, read the whitepaper. It describes the process in detail and even provides sample data and code for reproducible results.
Relational database administrators are often interested in measuring the resources a statement or query consumes. An example is CPU: we would like to know how much CPU time a query uses. Likewise, if a server’s CPU utilization is high, we would like to know which queries use the most CPU time. Query resource consumption is not possible to measure directly in many cases, and not always desirable even if possible. This paper demonstrates how a variation of linear regression, which we call weighted linear regression, can reveal the relationship between metrics such as query execution time and almost arbitrary metrics, both inside and outside of the RDBMS.