Unanswered: benchmarking - repeatable query execution time
I am trying to find a way to run a query more than once without it becoming faster. I have tried various things, such as: not using the query cache, flushing open tables, and restarting the server. These seem to help, but they do not solve the problem. Often the difference is still an order of magnitude. I suspect that it may be the key cache (is there a way to flush this?) and/or some OS factor (like pages being stored in memory). Does any one know how to repeat a query and have it be handled as if it wasn't already run?