I was reading an editorial by Denny Cherry entitled, “How much is that query costing”. Denny expresses the cost of a poorly tuned query in terms of the cost of lost productivity. Extended out for a whole year, tuning a query by as little as .4 second can possibly cost thousands of dollars in productivity.
Usually we determine if something needs to be tuned based on the hardware it takes to support an application. Denny reminds us that there are other considerations contributing to the cost and effort of tuning.
So, when do you determine if something needs to be tuned? Do you simply tune something simply because you think you can make it work faster?
Generally we get something to work, and make it solid. In a perfect world, we also have unit tests and integration tests. Then once things are nailed down we can circle back and tune things because our tests save us a lot of time confirming that we haven’t broken anything.
What are your thoughts on tuning? How much? When to tune? How do you know when you’re done? Share your thoughts here online or drop an email to btaylor@sswug.org.
Cheers,
Ben