I was asked this several times at TechEd so I'd like to answer it publicly. It's kind of complicated to explain so bear with me as I do it.
Earlier versions of SQL Server (for example, 6.5) had their problems but from 7.0 the storage engine has gone from strength to strength and is rock-solid. Yes, of course, there's the odd bug - I'm not going to pretend that we're perfect and I'm sure some smart-guy will post a comment pointing one out (thank you in advance) - but the simple truth is that the software is not the problem.
Now, ask yourself a question. Hands up how many people run their SQL Server on hardware? Anyone?
Daft question, right? Everyone runs on hardware. By 'hardware' I mean everything underneath the OS. There are millions of lines of code running in filter drivers, device drivers, controller firmware and drive firmware. There's memory, cpus, motherboards, controllers, drives, IO caches, SCSI cables - any of these can fail and I've seen them all, as I'm sure most of you have too.
It's the hardware that you have to be concerned about - with the myriad possible ways to setup a system to run SQL Server from combinations of hardware, who knows what could go wrong.
Actually, help is at hand. I realized as I was writing this that its the perfect plas to plug our new Always-On hardware verification. This isn't any new technology from us - I'm not trying to sell you something or force meaningless marketing on you, it's actually very helpful to you. In a nutshell, we've published the requirements of the SQL Server IO system and some hardware vendors have self-verified one or more of their hardware solutions against those requirements, publishing a detailed whitepaper explaining how they did it. This isn't any kind of certification from Microsoft, but it does give you more confidence that these solutions have been rigourously tested to the standards that SQL Server requires. Check out the
page for more info.
Ok - I kind of got distracted there - on to the post I
wanted to get out this morning...