Now that I've spent a significant amount of time thinking about Thin Provisioning (If you consider the amount of time it takes to fast forward through the commercial breaks in an episode of Heroes to be significant), I've a few thoughts about it.
First off I don't dispute the value of thin provisioning and I don't think anyone else will either, but there are more than enough pitfalls to ensure that it doesn't/shouldn't end up everywhere. While I'm sure over worked array jockeys will find it a godsend for quickly provisioning extra storage, (though I'd argue bringing the bind time of a LUN down to the 60 seconds you can do it with Fast Bind on FLARE 24 might achieve the same thing), Tier 1 applications would probably be best to avoid this. Lets face it, between something like Oracle writing metadata all over the place when you hand it a volume and some of the RISC UNIX OSs being generous with superblock backup placement if you provision 2TB of storage chances are they'll grab enough of it to ensure it ends up allocating all of it.
I also have questions around block level replication. Can it be done to non TP'd volumes without enforcing the equal size or larger rule? If you provision that 2TB LUN and use only 10% of it do you still end up replicating to a physical 2TB LUN on your destination side? Is it just a bad bad bad idea to replicate the contents of thin provisioned LUNs with anything except file level replication?
Interesting functionality, but ultimately this year's check the box feature. What's a bit more interesting is that NetApp are offering de-dup of primary storage. What's interesting about it is not that they're offering it but that so few people appear to care that they're offering it.
