SQL_Enthusiast (2/21/2013)
Bhuvnesh,Thank you for the response.
why you think that random cant be unique ? sql server internally generate this 32 bit alphanumeric code.
I didn't mean random like, "What is my eight year boy going to do next?" That is truly random. I meant, there has to be a rhyme or reason (aka logic) behind how to "randomly" generates the 32 bit alphanumeric code. Does anyone know the logic on how it builds that character string? Does it use the MAC address, date, time, tap into my deepest thoughts and encrypt them :w00t:
Older versions used to tie to MAC addresses, I think, but I believe the later versions use a pseudo random seed (for all intents and purposes, this is considered pretty similar to "real" randomness) rather than anything that's specifically tied to a specific machine. You may ask how that's guaranteed to be globally unique, and in the strictest sense, it's not. But you have to comprehend just how big a GUID is and just how low the chance of collision is.
It's a difficult thing for a lot of DBAs to accept as we're programmed to want everything to be guaranteed, regardless of the odds, but in reality, there are all sorts of logical guarantees in SQL Server that at extreme probabilities are violated. E.g. any kind of table constraint could be violated by hard disk/RAM corruption that happens to corrupt in a logically consistent way. CPU malfunction could cause calculation errors etc.
The odds against the collision of any two GUIDs generated randomly is in a similar or higher order of magnitude to these examples.