Comments on: More thoughts on NetApp, EMC, and Data Domain

By: ad

ad — Tue, 16 Jun 2009 20:42:45 +0000

Steve,
Please post on the antitrust issues. You’re our only hope.
—-First of all, I can’t be anyone’s only hope – for anything. Second, I don’t really get the anti-trust issues – I don’t really see any. Therefore, give me something to talk about and I’ll be happy too! — Steve

By: SFB

SFB — Tue, 16 Jun 2009 18:49:33 +0000

Steve-
Do you think the government will try to block the EMC-Data Domain combination. NetApp thinks there is a lot of problems.
http://blogs.netapp.com/jay/2009/06/deduplicating-customer-choice.html
—–I’m no lawyer but I have trouble finding any real monopolistic claim for either party here. It’s not as if there aren’t a billion other solutions that sort of do de-dupe. I like NetApp’s aggressiveness on the subject, but need to understand more to convince me it could stand. Maybe I’m just not seeing the big picture. Great question. – Steve

By: Tom

Tom — Fri, 05 Jun 2009 18:36:31 +0000

This deal is 100% about EMC stopping NTAP. NTAP had to understand that DDUP traded at $24 a year ago, so their $25 bid, while a high premium, was very likely to go up if EMC or anybody else piled on. Even $30 a share is only 25% over the year ago price. They had to know this was liklely to happen, and that a $57M breakup fee wouldn’t stop anybody. If they didn’t see it coming they need to fire their banker and their M&A guy. EMC jumped in not for any of the technology considerations mentioned here, but just to stop NTAP. Why? Because they can…
The sad thing is that nowadays EMC is dancing to NTAP’s tune, not just in NAS, but across the board. This is like the elephant that’s afraid of the mouse. Years ago EMC was afraid of Veritas at every turn, and now that looks absurd. Reacting to NTAP like this is only slightly less absurd. Fact is NTAP is a great outfit, but EMC dwarfs them. And NTAP, for all their positive attributes, ain’t exactly the greatest acquirer in the world. Odds are NTAP buys DDUP, and it evaporates in a couple years; I bet the integration goes sideways, the technology get’s old, the people leave, and EMC uses it’s three other dedupe products to beat them in the market. M&A is a low percentage game and NTAP is a weak player. So EMC doing this is pretty dumb in my mind, and diminishes them. Great leaders don’t react like this.
Fact is, EMC’s M&A program is broken and has been for a few years. Iomega?? That was their last big deal, and that was completely nuts. Before that you have to go back to VMware. EMC used to be a great acquirer. But all those people are gone, and now EMC hires a new M&A guy every year and buys odd little system management companies and goofy stuff like Iomega. We’re in the best “buy stuff cheap” market since 01, and they’re overpaying for stuff they already have, just to stop NTAP. Go figure.

By: W. Curtis Preston

W. Curtis Preston — Fri, 05 Jun 2009 03:45:43 +0000

@ Charles Hockenbarger
I may be wrong, but I didn’t see him saying that IBM will come out with something that targets the Symms themselves. I felt he was saying that that they should get a dedupe target product (which is the subject of the post) that is aimed at that market (i.e. the really big datacenter). Right now, the IBM Protectier is twice the speed of DD’s fastest box, but it could be much faster if they go from their current two nodes to more nodes. HDS is pretty much product-less in this space since the IBM acquisition of Diligent. Hopefully they’ll straighten that out soon. (Hint: both FalconStor and SEPATON are available and both have the throughput and capacity to reach the Symm market. Why doesn’t someone acquire THEM?)

By: W. Curtis Preston

W. Curtis Preston — Fri, 05 Jun 2009 03:33:02 +0000

I have to agree with Mr. Marks here. First, I don’t know where you (and others) come off saying that DD is 6-9x just buying plain old disk. Yes, attaching DD to plain old disk makes it cost 6-9x more. But then you have dedupe. If you get 10:1, which is a very conservative ratio (IMHO), you’re back below the price of the original disk without it. If you get anything beyond 10:1, then it becomes significantly cheaper than the cheapest RAIDed disk out there.
To your question about why we did this in the first place, we did it because tape has really started sucking as a backup medium. It’s nearly impossible to keep an LTO-4/T-1000/TS1120 tape drive happy when backing up directly to it. You HAVE to stream it by doing D2D2T. Then people liked disk so much they wanted to put MORE on disk, but it was too expensive. Then dedupe came along, making disk affordable again.
And with dedupe came the idea of being able to replicate your backups — and you simply cannot do that in most environments without dedupe. (That is, unless you throw out the whole works and go with something like CDP or near-CDP, which I’m totally fine with, but the backup world moves in very slow increments.) Replicating your backups allows you to have an onsite and offsite copy without touching tape — beautimus. No more worrying about encryption and the loss of a tape needing to be reported on the news. Data’s onsite and offsite magically every day without anyone having to move it around. If you want to make a tape copy, do it at the remote site and never move those tapes — again no need to encrypt.
As to someone else’s comment, if EMC thinks they’re buying DD to get dedupe on primary storage, I’d say good luck with that. I hope they have better luck with it than they did trying to leverage the Avamar technology and put it inside the VTL as an answer to Data Domain. MAYBE they can get the really smart DD guys in a room and start from scratch, but like I said: good luck with that.

By: Lou

Lou — Thu, 04 Jun 2009 20:30:24 +0000

There is a scenario where DD doesn’t win regardless. Netapp withdraws because of the EMC bid. EMC decides to withdraw as well. Stock price plummets and then Oracle (or some IP litigation firm) swoops in buys it all at $5/share and then starts asserting patent rights.
—-Ah, the lawyer card rears its ugly head! NetApp currently can’t withdraw – they have a definitive agreement. Only if EMC bids again can NetApp get out at this point. If somehow DD decided to walk from both, your scenario is viable, at which point the stock price would certainly plummet. It probably won’t plummet to $5/share based on this deal falling apart however – so some other things would have to happen for that to occur, in which case the vultures will certainly start circling. —- Steve

By: InsaneGeek

InsaneGeek — Thu, 04 Jun 2009 13:34:13 +0000

Here is my outlandish wonder… does EMC/NetApp have a larger range thought that instead of just traditional VTL that it’s about inline of primary? I’m guessing that inline dedup is pretty well locked up via patents. Would this allow them to move to inline primary storage dedup, rather than just post only? Having a inline primary storage method dramatically increases the value to where it’s only a moderately insane price rather than OMG crazy insane.
—-It’s a fantastic question! It is clear to me that de-dupe technology belongs, and will make it, to every step of the data contiuum – eventually. It is clear that the closer to the point of origin, the higher level of de-dupe occurs, and the greater downstream benefits can happen. NetApp absolutely knows this – which is why there overall success has been less precidated on pure capacity verses many others. It’s why they can maintain the margin structure they do. So, if either realizes that the key to “storage” moving forward (at least the key to extracting value $$$) is “less” is better, then there could very well be something to your argument that I hadn’t considered fully.
Having said that, if it were true, why not buy StorWize first and propogate from the other end of the spectrum? If “Primary” de-dupe occurs ubiquotously, I would argue there will be no need for secondary (backup) dedupe in the future.
Excellent thinking! —–Steve

By: Howard Marks

Howard Marks — Wed, 03 Jun 2009 15:37:16 +0000

Steve,
While most of your arguments make sense we of the chattering class have to stop comparing apples to oranges using the cost of a bare drive per TB against the cost of a DD box.
You maybe able to buy 1TB “nearline” drives for $169 at NewEgg that’s not really the cost of 1TB of storage. It’s the slot costs that make up most of the $$. EMC’s NY State price list, and don’t get me started on how badly the Albany bureaucrats negotiate, lists a DAE with 15 1TB drives at $28,414. That’s just a shelf no controller Etc. Even if a non-idiot could pay $19,000 that’s 7-8x the Newegg price. Total cost of a CX3 will be more like 10-15x.
For a backup target someone without an attachment to name brands could get an Infortrend 16 1TB drive array for about $7500 more like 3x bare drive cost.
Remembering that dedupe is really a software function the question is how much is that software worth to save cost of hardware, power, datacenter space and management overhead. I don’t have DD pricing handy to do a quick comparison but CommVault charges $3000/TB assuming 10:1 data reduction that’s less than the cost of a TB of Infortrend let alone Clariion.
- Howard.
PS: Looking forward to chatting at BD next week.

By: Charles Hockenbarger

Charles Hockenbarger — Wed, 03 Jun 2009 02:42:19 +0000

Steve:
How about you actually READ all of the comments on Chuck’s blog? Granted, I may be a small voice, but I’m not exactly sure how you could characterize my comments on Chuck’s blog as negative?
I see this as a great win for EMC. Expensive? Yes, but a win. I don’t necessarily disagree with some of your analysis here, though. I think you make some great points, and I need to go back and read all of it again. I think the issue of Donatelli is one that is hanging out there and could make things interesting for sure.
I’m not sure how exactly you’re expecting IBM (who has no product) or HDS (who’s parent company is too busy trying to make earth-mover and television building profitable) to compete with the V-Max. I am questioning some of the things EMC did with positioning the product, but it’s a smoking fast box with some huge potential, especially when FAST is released and they begin to phase out FC drives.
What, is IBM going to kill it with XIV? Is that a joke? Is HDS about to come out with the Hurricane and sweep them off the map? I mean, neither of these guys had arrays that were adequate competition for DMX-4, much less V-Max. I don’t see HDS acquiring any of these players making a difference. Has IBM’s acquisition of Diligent just stunned EMC and frozen them in their tracks with its outstanding execution and game-changing deployments? Ummm, not that I’ve seen.
I could be wrong, though.

By: Luddite

Luddite — Wed, 03 Jun 2009 00:02:01 +0000

Thanks Steve – thoughtful analysis.
What happens when dedupe and flash are combined? Does it offer a viable alternative to rotating media for primary storage?
Some data: Intel 80GB @ $650 = $8.12 / GB, WD 2TB @ $300 = $0.15 / GB
So flash is ~ 55x more expensive.
if seek time were zero, wouldn’t dedupe be suitable for primary storage?
If we get 20:1, dedupe improvement, double the data density (still around 35% for open systems as I recall), factor in energy savings, consider the server opportunity savings (increased server utilization when I/O delays reduced), then the economics shift.
RAID 6? Why if RAID 5 flash rebuild is fast. RAID 5? Not clear it’s the best approach either. And probably some other potential savings too such as compression.
Sure, not all data dedupes well. But it looks like enough could that deduped flash becomes a credible, cost effective primary storage consideration.
QTM licensed the Rocksoft dedupe IP to DDUP. Seems like EMC could use the dedupe IP for primary storage. I image EMC has a right of first refusal for QTM too. But if EMC buys DDUP, then it’s not clear that a QTM buy is important – other than to keep the IP away from others.
EMC’s purchase of DDUP could deprive competitors as you said, and exploit the combined DDUP & QTM IP for other purposes.