The database that I have inherited is infested with GUIDs. I should probably at this point inform you that I am not a fan of GUIDs, never have been and probably never will be especially after having to work with this schema.

The original reasoning behind choosing GUIDs as the clustered PKs was reasonably sound, to provide uniqueness to values that are generated at divers locations. However, due to the nature of the system with a high record insertion, the GUIDs are causing performance issues, locking up resources while page splits occur, blocking out waiting processes and ultimately resulting in timeouts.

Sure we could increase the timeout duration, but, to be honest, the wait time involved is unacceptable in our particular system, and not just from our perspective but most especially from our customers who do not appreciate delays.

You mention about hot-spots caused by using identity keys, well I've never yet had any issues with those, and I have been able to dramatically improve performance by eliminating the GUIDs in favour of identity. This on SQL 2005 as well.

However, the article was definitely thought-provoking and I did in fact learn something new, which is always a good thing, and for those of you who like using GUIDs, all power to you, I'm just not about to be a convert.

Mike