|
GUID = Globally Unique IDentifier. It's a little bit like an identity column, but without the distributed bug. What that means is, instead of going 1 -> 2 -> 3 -> you can guess what comes next when another database server is also going 1 -> 2 -> 3, etc, producing clashes, we get a random value. Something like 84D16AF9-6F2C-4EB3-9721-E20746A6A838. There are 32 hex digits, plus 4 dashes.
That means for storage purposes, the column is 16 bytes wide. That isn't morbidly obese, but it would never be featured on the cover of Playboy, either. One particular consideration is that if you have a 4 byte int value, that's 32 bits - just like the CPU and OS on a lot of production DB servers. It's long been speculated that this match is the best possible performance.
So, there are 2 ways you could use a GUID column. It can be a key, or it can be metadata. I'm not asking about just storing them - I'm curious what people's experience has been using GUIDs as keys for DRI, for searching, and so on?
|
