I love this quote from Frank McBath’s TechNote on Poor Disk Performance, Time Outs, Database and the SQL Server Errorlog:
You are at a party. There is plenty of beer, but the line is long. How do you fix the problem? Hire more bartenders.
Databases are exactly the same way. How do you increase the IO? Buy more disk and HBAs.
Unfortunately the answer is typically, “hire a faster bartender” as the customer tries to throw faster hardware at the problem. This parallels a much earlier statement by the late great Rear Admiral Grace Hopper known as the “oxen story”:
In pioneer days they used oxen for heavy pulling, and when one ox couldn't budge a log, they didn't try to grow a larger ox. We shouldn't be trying for bigger computers, but for more systems of computers.
I learned a ton about how SQL I/O stall conditions can effect performance from this paper and some of the references. I have been on a never ending quest to crack the ideal configuration for ILM implementations for the last 2-3 years and this is a big step in the right direction.
Also, check out Frank’s blog here:
Database+Disk+Performance