Inside Microsoft® SQL Server™ 2005: T-SQL Querying

Preparing Sample Data

When conducting performance tests, it is vital that the sample data you use be well prepared so that it reflects the production system as closely as possible, especially with respect to the factors you are trying to tune. Typically, it’s not realistic to just copy all the data from the production tables, at least not with the big ones. However, you should make your best effort to have an adequate representation, which will reflect similar data distribution, density of keys, cardinality, and so on. Also, you want your queries against the test system to have similar selectivity to the queries against the production system. Performance tests can be skewed when the sample data does not adequately represent the production data. ...