Thought I'd be able to ease back into the new year slowly, but - no - right off the bat IBM made a bold move that's got everyone talking.
You can see the press releases here and here. Basically, IBM paid somewhere between $300m and $350m for a storage startup.
I'm sure lots of people will weigh in here, but I'd like to offer a few thoughts on this one.
What's This Really About?
XiV (like all storage startups) has been on EMC's radar for a while now, just like all the other storage startups that are out there. But there's a bit of history here that some people find interesting.
The key investor behind XiV (a small Israeli startup) was none other than Moshe Yanai, one of the prime forces behind the Symmetrix, which itself has been arguably the single most successful storage platform of all time.
But the Symmetrix was designed in the early 1990s, and -- so much has changed in the 15+ years since -- so, for me, that's not what this is about.
It's about IBM and storage, pure and simple.
Is This The Next DS8000?
Andy Monshaw paints a picture of XiV being about something new, e.g.
"The acquisition of XIV will further strengthen the IBM infrastructure portfolio long term and put IBM in the best position to address emerging storage opportunities like Web 2.0 applications, digital archives and digital media."
When I saw this, I thought they'd either made an incredible (and expensive) mistake, or it's a bit of misdirection on IBM's part.
Why?
Well, we know how XiV built their platform, and we also know what this kind of "new information" requires from a platform perspective, and it just isn't adding up in our estimation.
I could list about a half-dozen things in the XiV product that doesn't make sense for this market, and I could list another half-dozen things that XiV doesn't do that they'd need for this space.
No, I think we're getting a preview of the future replacement for the much-maligned and neglected IBM DS8000. Much has been written about how this product has fallen significantly behind the competition in this part of the market, so much so that it isn't even taken seriously in many circles.
And it's clear they have to do something about it sooner than later.
If you think back to the original IBM announcement, they were very proud of the fact that the DS8000 used IBM's pSeries server technology as a controller. Remember the LPAR feature of running applications on the array?
Well, if you're a server vendor, the idea of using your servers to build storage arrays is downright intoxicating (think about Sun for a moment), but this approach doesn't work out so well for certain parts of the market.
I'm sure part of the appeal to IBM is they saw the potential of doing this again with XiV, in addition to getting a modern feature set.
Go check out the XiV website for a moment -- and ask yourself some questions:
First, could this be a replacement for the DS8000? Yep.
And, second, if you were building a Web-2.0-type storage platform, would you build it this way? Nope.
Now, if XiV is intended to be the basis of a DS8000 replacement, it wouldn't do well for IBM to position it this way, regardless of its ultimate intent.
I'm taking bets here -- how many people want to bet against me on this one?
Is This Really About Dell?
Don't underestimate the domino effect in M+A. If your competitor does something big, you think long and hard about it. We saw it when EMC was doing lots of acquisitions, which tended to spawn all manner of follow-on by our competitors.
Dell just layed down big bucks for EqualLogic. My take is that they looked at their server business, and realized that they had to own some of their own products in this space. And they spent big in an attempt to do so.
Now, if you're IBM, how do you react to this big move by an arch-competitor?
Let's see -- for IBM, at the entry level, you've got the LSI stuff. In NAS, you resell NTAP. In the large midtier array market, the DS8000 is about 3-4 years overdue for a refresh. And, architecturally, you don't have anything at the high end to respond to EMC and HDS, despite the inevitable protests to the contrary.
The only interesting storage IP you really own is SVC, and I don't think that's not enough to float a mighty ship like IBM.
Maybe IBM felt the need to make a Dell-like move?
Anyone want to bet against me?
Is This Really About IBM's Execution In The Storage Market?
I think it's fair to say that IBM's execution in the storage and information management space hasn't been all that stellar.
Yes, I give them credit for trying, but their success rate hasn't been exactly phenomenal, and, of course, no one is perfect. And, one thing you've got to give them, they sell the heck out of whatever's in the price book -- no one does it better.
In an evil moment, I considered that the significance in the XiV name was "fourteen", and I tried to make a list of the thirteen not-so-good IBM storage ideas that preceded it, but I thought that was too harsh. Besides, once I got beyond StorageTank, SSA, IBM storage LPARs, storage bricks and a few others, I could only come up with a half-dozen or so for my list.
It's great that they're acquiring technology where they need it. Hey, look at what EMC has been up to for the last few years.
The real trick is making it work, and coming up with an offering that makes a difference with customers. Best of luck, guys, you're up against some stiff competition.
And I think that will be IBM's biggest challenge, won't it?
This has nothing to do with Dell and has everything with having a product to sell to existing IBM customers or use internally for IBM Global Services that is 'good enough'.
Remember, customers do not always buy 'best of breed' and IBM customers for years have invariably purchased 'good enough'.
IBM are unlikely to sell this product into customers that are not already buying IBM Z-series and p-series servers. They don't have a direct workforce strategy or a channel strategy that would support this.
I do expect however that IBM will make some inroads into the z-series space that IBM has competed on - and struggled to maintain commanding share - with EMC and HDS. They now will have something other than the DS8000 that they can bundle - and once it is brought into the IBM corporate portfolio and blued, it may be 'good enough' for customers to make a 100% IBM decision when laying out the cash for a new server-storage environment.
Posted by: mgbrit | January 02, 2008 at 04:31 PM
Hi -- the points on "good enough" and "bundling" are well-taken.
It's not clear to me that XiV has the architectural underpinnings to support zOS attachment, so that's a bit of an open question.
Thanks!
Posted by: Chuck Hollis | January 02, 2008 at 04:55 PM
In my experience from Symmetrix, supporting Mainframe connectivity is a pretty challenging task. Besides the complex and archaic CCW protocol on both Ficon and Escon, you need to be able to store CKD data efficiently and allow random access to any record in a track. Once you have basic IO going, you need to support numerous applications: PAV, Dynamic PAvs, Dynamic Path Reconnect, GIDN locks, XRC, multi-allegiance, PPRC, FlashCopy, and many more. IBM, bless their soul, keeps coming up with new storage protocol changes frequently, and the list from the last few years is staggering. In other words, I believe that even if XiV decides to implement MF connectivity, it will take them years before they have something close to the compatibility level of HDS, EMC or DS8000.
What this means is that IBM will need to continue to sell DS8000 and XiV, two completely incompatible architectures, with different management systems, different user interface, different performance characteristics, and of course, different hardware. If I were an IBM customer now with a mixed MF/OS environment, I'd get pretty nervous.
Posted by: David Meiri | January 04, 2008 at 03:04 AM
You bring up some very interesting thoughts.
Yes, mainframe storage is very different and very hard, but Moshe has been there before, as have some of his engineers, and -- unlike our experience at EMC -- he'll be working at IBM, or that's the theory anyway.
Going a bit farther, you're right, we might see a replay of what happened in the AS400 (now iSeries) market where IBM succeeded in closing off most alternatives to IBM storage.
I could see the current DS8000 as the "lock in" array for mainframe environments, mostly through the continual feature enhancement IBM puts in zOS.
The bottom line: customers lose. We talk to iSeries customers all the time, and they don't like being trapped by IBM.
The DS8000 can't make it as a mainstream storage platform -- it's toast -- so the XiV platform would make a reasonable alternative for most of the workloads IBM encounters with their xSeries and pSeries servers.
Thanks for writing!
Posted by: Chuck Hollis | January 04, 2008 at 07:09 AM
What I don't get is why you guys keep going on about DS8000 being dead? Its just has one of its best years ever sales wise, and other than the concerted effort of spinning obvious FUD, I don't see the correlation. It is being invested in, it sees a substantially larger investment year on year than we do in SVC and the roadmap continues.
CKD will always be the thorn in an Enterprise Controllers side, but maybe there's a new way to think about it.
Posted by: Barry Whyte | January 04, 2008 at 03:29 PM
I know you're going to find this hard to believe, but -- this ain't FUD spinning -- we really believe that the DS8000 is dead from a product planning perspective.
We see it falling farther and farther behind EMC and HDS and few others in competitive features. We don't see how the dual-controller architecture will support future requirements, especially for large enterprises. We don't see it picking up decent modern technology, like 4Gb FC, larger drives, IP connectivity, etc.
Just like in the real world, if it doesn't look like it's moving, we assume the worst.
Trust us, we LOVE a good competition! But we see that IBM picking up their toys and going home on this one, unless there's something really big coming, like XiV/Nextra.
Now, that being said, IBM does an excellent job of selling / bundling / moving whatever they've got in their price book, so I'm not arguing "best year ever", even though IBM doesn't provide y/y figures or anything like that.
Lots of people buy storage at the same time they buy everything else, and that plays well in IBM's favor. Every EMC storage product has to be sold largely on its own standalone merits. I'd be curious how much DS8000 (or any IBM storage for that matter) gets sold "off base", e.g. outside the context of a larger IBM integrated deal.
To be honest, the Symm folks are watching HDS much closer than IBM these days ...
CKD is always a problem, and alway probably will be. It'd be nice to see an open, standard and well-supported storage interface architecture in zOS and iSeries, but I don't think that's in the cards.
Oh well.
Happy New Year, Barry!
Posted by: Chuck Hollis | January 04, 2008 at 04:28 PM
Excellent points there Barry!
Coming from a largly IBM mainframe base (MVS/OS390 and zOS), and then moving into the server arena - with EMC/IBM/HDS equipment - SVC's included, it is very true that IBM mostly gets their storage deals done via the z/OS picture.I really only see HDS/EMC as players currently - (Although my son will vehemently argue with me on this - but then he does work for Big Blue ...).
It will be very interesting to see what they come up with, and where/how they position the XIV. Might be another RVA story - only the next 12 months or so will tell...
Posted by: Christo Heuer | January 07, 2008 at 09:03 AM