Speaking of which, Symmetrix storage provisioning has been improved from nearl 1hr (elapsed) per TB allocted to less than 8 minutes/TB in 4773. Virtual Provisioning reduces this even further for the actual provisioning, even if used to alloacte so-called “fat” devices (all fully allocated from the storage pool). In this case, you’re not necessarily using less – or more – storage, just getting it allocated faster. For some, worth the price of VP alone.

If your array is running at low capacity utilization, you probably do not need to spend a lot of money for thin-provisioning.

But then, many storage arrays and applications today run at low capacity AND low performance utilization – these are the prime candidates for TP.

IMHO, the conspiracy theories are just FUD: pricing and promotion won’t be what inevitably limits the demand for TP, though.

No, as with every TP implementation on the planet, **by definition** the trade-off of performance vs. capacity utilization, along with the inherent risks of over-subscription and uncontrolled rate of consumption will be what slows TP implementation – regardless of vendor or platform.

Undoubtedly, many customers who understand these implications will decide NOT to use TP for their Tier 1 applications – at least not until they are sure they understand the implications and are prepared to deal with them.

Bottom line: TP is not a miracle solution to increased capacity utilization for every application. It won’t be adopted universally overnight simply because it is now available on Platform X. Like RAID 5, adoption will be gradual, and thus the impact on capacity demand will ramp slowly over the coming years – probably to go unnoticed against the backdrop of the continued 60% CAGR demand for capacity.

As to the nail in 3PAR’s coffin…I dunno, you tell me: When all your competitors offer the only differentiated feature you had, AND they have a rather long list of differentiated features that you DON’T have, plus a massively larger installed base of satisfied customers – what then? I’m thikning that things probably start looking pretty bleak, leaving the one-time pioneer defenseless except for maybe “my implementation of TP is better than theirs.”

No wait – hasn’t their CEO already started playing that song?

If an application needs the performance that 60 15K FC drives delivers but only 30% of that capacity, you can’t buy 30 drives and use TP to get 85% utilization. The performance just won’t be there. (my numbers may not add up there… I’m winging it here…).

I think Tucci is saying that it won’t impact EMCs bottom line because they will offer it but they won’t promote it. It will allow them to “check the box” when companies evaluate their products (like NQM). Also, Tucci knows that if people implement TP incorrectly (very easy to do) they will be forced to buy more disks from EMC to get the performance you need.

I do think that when EMC finally announces TP it will put a nail in 3Par’s coffin. TP is about the only thing 3Par has that’s relatively unique.

That was the article that I had fun with regarding IBM’s insistance that Flash was unnecessary and that they planned on pushing tape instead (in this post http://thestorageanarchist.typepad.com/weblog/2008/02/0068-rotflmao.html ).

Still, I don’t think IBM has ever actually said that they’d be delivering thin provisioning on the DS8000. Andy said only that “this is the year we need it.” And even Clod Barerra didn’t actually say it was actually coming on the DS8000 – what he said was that it was a “good expectation” (as I discussed here: http://thestorageanarchist.typepad.com/weblog/2008/04/0076-oops-ibm-d.html )

Somehow I’m starting to think that IBM is being very careful with its choice of words, possibly because they plan on delivering thin provisioning on a new hardware platform instead of on the DS8000 (the rumored DS10000?). If so, that would put them in the same camp as Hitachi, forcing customers into a hardware upgrade to get new features – as compared to the Symmetrix Virtual Provisioning, which is avaialble for both DMX-4 AND the prior-generation installed base of DMX-3 arrays.

But at least now we know that the SVC is going to offer thin provisioning. Wonder if it’ll be before, or after, it is available on the DS series…

My article on the DS8000 thing was linked on the word “heard” above, but it prob. got lost with that short an anchor text:

http://searchstorage.techtarget.com/news/article/0,289142,sid5_gci1302575,00.html

But I’m not sure where you heard IBM say that they were doing thin provisioning for the DS8000, though. Clod Berrara was quoted as not dismissing the idea, but I haven’t seen anything near as concrete as what BarryW said about SVC and “space efficient volumes.”

Maybe I missed it.