Professional Microsoft® SQL Server® 2008 Administration

2.4. Burning in the System

Before you move a system into common use, you should "burn it in." This means that you should stress the server. It is not unusual for a server to work in production for months or years with a hardware problem that existed at the time the server was deployed. Many of these failures will not show up when the server is under a light load but will become immediately evident when the server is pushed hard.

Several years ago, using SQL Server 6.5, I would get a 605 corruption error, which forced me to restore the database that had the error. When the restore was complete, I'd run my DBCCs and get a clean result. Within a week I'd get another 605 error and have to stay up late in the evening restoring the same server. After significant frustration with this recurring problem, I discovered that I had a disk controller failing only near 100 percent utilization. This is exactly the kind of problem that burning in your system is intended to expose prior to production. It's much easier to fix hardware before it goes live.

A second use of some of these tools is to compare the I/O throughput of your new system to your requirements or to the performance of an older system. While many tools are available, you can get two of them from Microsoft. One tool is called SQLIO, which is a disk subsystem benchmark tool. Microsoft provides this with no warranty and no support, but it is still a good tool to use. It is available from www.microsoft.com/downloads; search for SQLIO. ...