EMC’s XtremIO product is launching next week on November 14. This is the beginning of the end for EMC’s Symmetrix line and the toughest product transition they’ve ever attempted.
You may recall last year when StorageMojo decrypted the content of a Storage Newsletter interview with Chuck “Pravda” Hollis, EMC’s marketing spokesman.
After noting that the pre-announcement was meant to freeze the all-flash array market while NOT positioning it as a Symm replacement, StorageMojo opined:
It won’t ship until the 2nd half of 2013 at the earliest and may slip further. But an H1/2013 announcement is likely.
When EMC bought EIO, they predicted a Q1 2013 launch. At the time EIO hadn’t shipped a product.
Ever the optimist
StorageMojo was chided by Mr. Hollis, who denied any problem with ship dates:
I was very dismayed yesterday to see that Storage Newsletter made a hash of a recent interview. You can tell by reading it that they had multiple and severe translation challenges. And, of course, they never bothered to send along a draft back to me for verification — which is standard practice in these cases.
We’ll be talking with them about that.
I was doubly dismayed to see that you had grabbed a snippet of their poor work, and decided to make it all your own.
To be clear: the Xtremio product is doing quite well, thank you, and meeting all of our expectations. We’re looking forward to a public GA in 2013.
The energetic publisher of Storage Newsletter, Jean-Jacques Maleval, denied any transcription errors. Given that the interview was conducted in English – unless Chuck is fluent in French – there wouldn’t have been any translation errors at all.
Chuck also said that the product was in “early access” – beta – last December. Then in July EMC President and COO David Goulden said they were on track for a Q4 announce. EMC offered the product on a “directed availability” for certain use cases at that time.
Veterans know that a long beta is a sure sign of a troubled program. But let’s put all that behind us. Where are we 11 months later?
The StorageMojo take
Given the success of all-flash arrays from Kaminario, Nimbus, Pure and Violin, there are three major concerns in Hopkinton:
- They need a tactical product that they can – if needed – give away to keep the currently little guys out of EMC accounts.
- But they can’t let the XtremIO product kill presumably larger and just as profitable Symm sales.
- Also, it would be nice if the product didn’t blow up customer operations.
My, that is a conundrum!
If I were Violin et. al. I would order up 25,000 coffee cups with my company logo on them and use every means possible to get them into EMC’s major accounts. Force EMC sales to lead with XtremIO at big discounts, making the price/performance gap between Symms and flash arrays all the more stark.
Despite a year’s worth of effort the XtremIO product is not as mature as ones shipping for a couple of years. EMC is caught in a bind of its own making – which doesn’t happen very often – and now is the time to make them as uncomfortable as possible.
EMC will trot out beta customers to tout the wonders of XtremIO. But a November 14 announcement – about the last possible day to get some reasonable attention in 2013 – makes me wonder how ready the product is. If you order one let me know when delivery is promised – and if it works.
Courteous comments welcome, of course. I’ve done work for Kaminario and Violin.
Mojo,
You make it sound like as though the sky is falling because EMC didn’t realease XtremIO in time. Btw, XtremIO went directed availability beginning of the year as promised.
Let’s level set, the flash market is still in its infancy. Not all workloads belong in flash! For example, Flash is designed for small block, high read workloads. Flash drives are still more than 5X the cost of spinning drives last I checked. I wouldn’t want my customer running on consumer grade flash either. Flash arrays also depend on data that is highly redundant, as VDI. For this reason and several others, Deduplication is key technology to reduce the amount of data stored on these arrays to reduce the cost, reduce the wear and tear on NAND cells, etc. If your workload is write intensive (think log files) or does not dedup well (think databases), what is an all flash array vendor going to tell their customers when the array they bought doesn’t perform or can’t fit all their data into the array? Me customer, buy another, and another. With EMC, customers have choice to place their workload on the appropriate storage based on their application needs, not a hammer looking for a nail. VMAX and hybrid arrays will not go away any time soon. XtemIO is worth the wait! EMC can afford to harden the XtremIO all flash array, and not have to release products too soon to appease wall street.
The vendors you mentioned all have good products, but they were released way too soon, with lack enterprise features of hybrid arrays like replication, and rely on storage technologies invented for spinning media to be leveraged in their arrays like RAID that cause write amplification, create the need for resource intensive garbage collection, degrade in performance as the capacity fills up, have to turn off features to keep up with performance because the active/passive nature of their controller design. Not to mention, many of the products you mentioned are not even scalable to support growth and rely on an active/passive controller does parity calculations in NAND SSD, rather than DRAM like XtremIO.
Let’s continue to level set here, IDC predicts the all flash storage market to grow to only $1B in 3 years. Last year it accounted for just $300M. The external storage business was $25B in 2012.
Gardner predicts that in the next 3 years, >80% of the flash companies you mentioned will also be out of business, because he hopes of being bought by large storage players like IBM, EMC, HP are long gone, as all of them have already bought the likes of TMS, Whiptail, and XtremIO. Where is the investment in coffee cups, and further innovation going to come from?
John (EMC employee)
I find myself agreeing with SOME of what John (EMC employee) stated. Flash does indeed have very specific use cases. The new kids on the block are lacking the maturity and features required for serious consideration. Probably 80% of them will be gone within three years. (Still, I guess Pure was drawing enough blood to justify the sue ball…)
Clearly EMC recognized the reality of Symmetrix being less and less competitive with 3PAR V, DS8k, and VSP. (Unfortunately for HP a lot of their own folks still understandably prefer a re-branded VSP as the air gets more rarefied…) The replacement timeline with XtremIO is also clearly taking a LOT longer than they want, but at least it is not a Spinnaker class fiasco.
So, where does that leave you if you need well proven Tier 1 and have some flash use cases? For the time being I would be least worried about VSP, but that is just me. Next Christmas is the earliest I would be willing to seriously consider EMC’s new ‘sheriff’; VNX2 is also going to first need a few more real world miles behind it IMHO.
Disclosure NetApp Employee – opinions are my own.
John … I get this old feeling of “here we go again”, when I read stuff like “Need .. resource intensive garbage collection, degrade in performance as the capacity fills up, have to turn off features to keep up with performance”
I’m not a particular fan of many of the flash start-up’s marketing claims, especially those touting “1 meeeliooon IOPS” for unrealistic or unspecified workloads, but … Dang … did EMC just dredge up their old anti-NetApp FUD to throw at the new flash players or are there actual solid proof points for those claims ? I think it’s right to say that there are some challenges to some of the performance claims when you start to layer what many might be considered to be real-world workloads, but to then claim that this is “because the active/passive nature of their controller design.” seems out of order. Are you really suggesting that if they’d been active/active that NONE of those issues would arise ? Active/passive architectures do indeed have some drawbacks compared to active/active and scale-out (e.g. the relative inability ability to absorb unusual workload spikes), but attributing all of that to that one cause seems somewhat misleading … why not just say “active/passive is BAD … hmmmkay ?” … I feel like its “find a deficiency and then attribute all possible faults of all competitors to that one deficiency, and then tar them all with the same brush” .. it just feels wrong somehow.
Then there’s this lovely little snippet .. “parity calculations in NAND SSD, rather than DRAM” … really ? I’m assuming you’re dumbing that down, because AFAIK, CPU’s tend to do the actual parity calculations from data in DRAM for all architectures. Yes with traditional RAID at some point in time you need to read the missing “stripe” data from the storage media and combine that with what you already have in memory, and an X-Brick may not do traditional RAID, so its possible that XtremeIO’s data layout algorithms _never_ needs to read missing “stripe” data, and that all data protection/ECC coding is _always_ done in DRAM. But that’s not what you said, and I’d be very surprised if there weren’t some reads that need to be done from flash at some point to make that work (even if it’s just currently uncached metadata that needs to be invalidated and written back to new locations)
Finally “are not even scalable to support growth” with another dig at the “active/passive controller architecture”. If you’re saying that a scale-out is great feature that many of your competitors lack, and that it can make performance scaling for a system really efficient, and furthermore elegantly solves a number of management issues when done well, then I’d have to agree with you. But the tone of your argument suggests something entirely different to me.
Believing in your products, and standing on the strengths of your solution is great, but XtremeIO is in many respects, a new start for EMC, it from my perspective, it would be nice if that new start was built on a foundation like this.
“We think of the customer, our tribe, ONE EMC, the community at large, the world – and never go negative on the bad guy”
Not the same old FUD practices of yesteryear.
Regards
John Martin
Great article!
Disclosure – Pure Storage Employee, formerly with NetApp
I’d like to congratulate EMC on the pending release of XtremIO.
No one debates is 1. Flash is exponentially faster than spinning mechanical drives and 2. Flash costs 5X of spinning mechanical drives. What is up for debate is the means at which one should consume flash.
SSD is a storage medium, what’s possible with that medium will be based upon the software / hardware design in which it is managed. If one positions flash merely a performance tier, then they are missing the greater picture. Flash can provide massive gains in performance, reductions in data center resources (i.e. rack space & power), and can leverage storage savings technologies (i.e. data reduplication & compression) to actually cost less per GB than spinning mechanical disk.
Kudos to EMC & the XtremIO team, while they are my competitor they understand the rapid pace at which the disk drive market is changing (or collapsing).
Cheers,
v
Hey Robin!
What’s up with the personal attacks? Certainly you’re better than that.
— Chuck
Chuck, what personal attacks? I think you said some insupportable things in your comment and I called you on them. That’s not a personal attack, that’s what I sometimes do with companies and, on occasion, with people.
After all, EMC had no problem threatening StorageMojo 6 years ago, as you may recall, on totally laughable grounds.
Disagreement is not an attack. Happy to continue the conversation.
Robin
Hi Robin
Chuck “Pravda” Hollis here.
I don’t know if this conversation is going to go anywhere, but I’m going to give it a sincere try — point by point.
First, the incident with Jean-Jacques was regrettable. He and his assistant admitted to transcription errors, and I later found out that I wasn’t the only one who had experienced that particular issue. Since then, Jean-Jacques appears to have modified his approach to “live” interviews, which is good for all.
Second, we all felt the incident six years ago was also quite unfortunate. EMC is a large company of 50,000+ people, and many of us stepped in to help rectify the situation once we were aware of what happened.
If you want to hold me personally accountable for what happened, that’s your choice.
Also, in a minor update, I now work for VMware, not EMC.
But — at the time — I presented the XtremIO product with the best available information, and — guess what! — it’s largely turned out as I described. The product will be announced tomorrow (yes, it’s still 2013!) and — from what I know — it will exceed expectations.
Like you, I’ve been blogging for quite a while with a modicum of success. But — unlike you — I’ve never felt the need to personalize industry discussions and vilify people.
To each their own.
— Chuck
mheh…
Robin,
Take all the vendor reps and put them on a long boat ride.
Encourage practitioner feedback and conversations otherwise the conversation will digress.
Chuck,
I simply asked you to identify one personal attack I’d made on you. You didn’t. But if you want to get specific please do. You’re the one saying “personal attack” and “vilify,” not me.
Also, your insistence that Storage Newsletter got the transcription wrong is not supported by the facts. Anyone can listen to the entire interview on an MP3 downloadable at the bottom of this page and see that the parts I quoted were accurately transcribed.
This is not a personal attack, this is called fact-checking – which journalists used to do daily – but has fallen into disfavor as the facts showed a liberal bias. You’re entitled to your opinion, but not your own facts.
To get technical, I subjected your original interview to a close reading and forms of textual analysis – most of which I learned studying the Bible in college – informed by decades of experience in engineering projects and competitive analysis to figure out what was going on with XtremIO project. I hope this didn’t create problems for you inside EMC because you were doing the best you could with the facts you had. I don’t see how you could have done better short of not talking to the media at all.
As for EMC threatening StorageMojo, I confess I found it an odd coincidence that 2 weeks after I rubbished some claims you’d made (see EMC: “These Aren’t The Droids You’re Looking For†I got a remarkably ill-informed – a 1st law student would have done better – letter from a 60-something EMC lawyer who obviously a) didn’t know how to use Google, and b) had acted with such haste that I concluded
And, as I thought at the time, somebody with the clout to get him to act in haste. No one has ever owned up to who pointed attorney Clark at me.
Chuck, if you think that playing the victim is a winning strategy, be my guest. In the real world you are considered a mouthpiece – and, I think, a damn good one – for EMC and now VMware. And with your 6 figure salary and your bully pulpit at EMC and now VMware, no one is going to feel sorry for you. If you think StorageMojo subjecting your online comments to critical analysis is a personal attack then we have very different definitions of the term.
Regards,
Robin
The article was lame but the comments were really fun to read.
Look, I am an EMC partner and reading all this reminds me that everyone in this thread is lost in the weeds. Not one of the points in this thread matter to 99.999% of the buyers and users of these technologies – sorry to be the bearer of bad news. You are splitting hairs over a teeny tiny workload that could probably be handled by nearly every vendor in the conversation (well…all except NetApp because I just don’t know when they intend to join the party). And in terms of delivering on time or not on time, again completely irrelevant in the long term plan of any customer buying this technology. We partner with EMC because they solve most, if not all, storage problems with our customers. Could XtremIO be defficient in 1 area that another smaller player excels in? Yes, quite frankly they should be. They are developing products to fit a storage strategy. Other storage vendors have a similar story and their partners should feel the same way. As long as EMC continues to deliver high performing, reliable and fairly priced solutions we will be here for the ride. Adding XtremIO to the portfolio is a welcome change, on time or not.