Forecasting MySQL Scalability with the Universal Scalability Law

Forecasting a system's scalability limitations can help answer questions such as “will my server handle ten times the existing load?” and “at what point will I need to upgrade my hardware?” Timely answers to these questions have more business value than exact predictions. Mathematical models can help reduce guesswork while avoiding the expense and time required for real load testing. Dr. Neil J. Gunther's Universal Scalability Law is such a model. It predicts a system's deviation from ideal scalability, based on simple measurements that are relatively easy to collect. In this paper we show how to model a MySQL database server's scalability, in terms of throughput, for two different servers and workloads.

November 2010