Just because it’s there; you don’t have to use it. This is a message that I’m trying to get across to people so that we can successfully implement Thin Provisioning and get some benefits.
Marc Farley, in a reply to my post on credibility, points out that you might not want to use Thin Provisioning for user-storage as it does not stop your digital data packrats just blowing away your space. Alex McDonald sensibly points to quota management; we certainly don’t allow any of our users to have an unlimited quota, even I don’t have an unlimited and I’m the Storage Manager. I *may* have a larger than normal quota for somethings such as email, there’s got to be some priveleges!!
But this post is not about that; we have been working on standard LUN sizes, it makes dividing an array alot easier and there is a fair amount of debate about what a good size for a LUN is; should it be 10 gigs, 50 gigs or 100 gigs. 1 gig is almost certainly too small, where 500 gigs may well be too large. And even 10 gigs can be wasteful if you only want 1 gig of space for some redo logs.
So I am trying to train/convince people; just because they have been given a 10 gig LUN; they don’t actually have to allocate it all and if they don’t allocate it all; my friends the DBAs won’t simply write nulls all over it. In the past, there was no real benefit but now if we go down the Thin Provisioning route; we can get benefit and I can keep those dastardly DBAs under control.
We can allocate some large LUNs and potentially never have to allocate another LUN, but we can do so knowing that if won’t be wasted if isn’t used and nobody is simply going to fill it up. The server admins simply have to increase the size of a volume/filesystem and never have to bother the sleepy storage admins who have been fighting real problems all weekend; such as multiple blade failures in the SAN, an FA going down and the fall-out from that (it was not a good weekend!).
Well put Martin. Thanks.