• 06/02/2015
    8:00 AM
  • Rating: 
    0 votes
    Vote up!
    Vote down!

Gartner All-Flash Array Magic Quadrant: A Bad Influence

The market analyst firm's Magic Quadrant for all-flash storage arrays is based on criteria that hamstring vendors and reduce customer choice.

It should come as a surprise to no one that Gartner, as the biggest IT analyst/advisory firm, has a lot of influence over what its clients, and less directly the rest of the IT world, buy. In recent conversations with senior executives of several storage vendors, I've become concerned that by arbitrarily defining market segments, Gartner is not only influencing what its client do buy, but also what products they -- and the rest of us -- can buy.

I've written before about Gartner Magic Quadrants and how technologists misread them as a product buyer’s guide rather than the general advice to Fortune 500 CIOs that Gartner analysts intend them to be. I frankly don't have a problem with the global CIO CYA strategy that moves a vendor up on the “ability to execute” scale if it has a support organization in sub-Saharan Africa.

My problem is with how Gartner defines a market I think I understand pretty well -- the all-flash array --  and how many vendors are so driven to be included in the analysis that they design products to fit Gartner’s definitions even when they believe that there is, or might be, customer demand for something else.

When publishing its first Magic Quadrant for all-flash arrays, Gartner made decisions on which products qualified that seem to be based more on the products' marketing than their capabilities. The problem starts with the fact that the folks at Gartner decided that to be an all-flash array, and therefore included in the Magic Quadrant, a vendor had to offer an all-flash array as a separate product with a unique model name and model number.

Learn more about the changing storage landscape in the Storage Track at Interop Las Vegas this spring. Don't miss out! Register now for Interop,  May 2-6, and receive $200 off.

Gartner even lists vendors that make “SSD” storage arrays that didn't qualify for the MQ such as Dell and NetApp, implying that a SSD array like  a Dell Compellent is somehow less than a real all-flash array like Pure Storage's or an EMC VNXF, even though the flash in those systems is packaged in SAA SSDs, just like the flash in a Dell Compellent SC2040.

As an IT architect, this makes no sense to me. When choosing the product (or products) I want to solve my storage performance problem, I don't care if they have a unique product name or share the name with a hybrid, or even all-disk, system using the same architecture. All I care about is that the storage system provides the performance and features that address the problem I'm trying to solve.

While the unique product name and number criterion was arbitrary, analysts do have to draw the line somewhere. Gartner’s inclusion criteria also state that the vendor must offer an all- flash array -- fair enough, it is after all the Magic Quadrant for all-flash arrays. However, I object to Gartner's insistence that the system “cannot be reconfigured, expanded or upgraded at any point with any form of HDDs within expansion trays via any vendor's special upgrade, specific customer customization or vendor product exclusion process into a hybrid or general-purpose SSD and HDD storage array.”

These criteria mean that EMC VNX-F qualifies as an all-flash array while the all-flash versions of NetApp’s flagship FAS don't, even though the VNX-F and all-flash FAS are as directly competitive, and comparable, as any two storage systems on the market.  Both are the latest versions of unified storage systems filled with SSDs and most storage architects that were considering VNX-F would also look at an all-flash FAS.

In talking to senior executives at several storage vendors, I've discovered that these frankly stupid criteria are not just keeping products off the Magic Quadrant, but also keeping more flexible products off the market. In order to meet Gartner's criteria, storage vendors are selling separate all-flash and hybrid storage products even when they run basically the same software.

The net result is that customers of Agile Storage who buy an Agile-F AFA and later decide they want to add a couple of shelves of 7200 RPM disks to hold their colder data are screwed. Agile has to refuse to sell them, or even allow them to install the spinning disks as a “customer modification.” That would make the Agile-F a mere SSD array, not an AFA, and push it off Gartner's AFA Magic Quadrant.

Who the heck -- and those of you who know me know that heck was not my first choice of words -- is Gartner to tell vendors they have to name their products a specific way, and worse, that they must reduce the flexibility they offer their customers? Today's data center is a much more dynamic environment than in the past and vendors should be boosting, not cutting back on product flexibility.


Agree 100%


You are dead on right here. I'm too am growing concerned about what Gartners says is "in" and "out", not only for the all-flash array market but just about every MQ they create. They are almost too diverse for their own good. We first brought this to light when they published their storage MQ, which I for one can't understand why that is separate from the AFA MQ.  See our tear down of the storage MQ by going to storageswiss dot com and searching for Gartner.



George Crump
Chief Steward, Storage Switzerland


Gartner even prefer's their own acronym?

I can remember my first Gartner report where they wouldn't even call the All Flash Arrays (AFA) by that term.  They opted for their own term - Solid-State Arrays (SSA).   

It's worse than just the MQ

After I sent in the post I saw that Gartner was reporting market share numbers for all flash arrays using the same ridiculous market definition they used in the Magic Quadrant.  As a result the numbers they reported

  • EMC at $444m — a 31.1 per cent share ($74m in 2013)
  • Pure Storage with $276m — 19.4 per cent
  • IBM with $233m — 16.4 per cent ($164m in 2013)
  • NetApp had $110m — 7.7 per cent ($71m 2013)
  • HP at $102m — 7.2 per cent
  • Violin Memory at $60.2m — 4.3 per cent
  • Solidfire roughly at $50m — 3.6 per cent
  • Kaminario roughly $50m — 3.6 per cent

impled that Dell, HDS and the other vendors who make products to best serve their customers, not to fit in Gartner's definition of an AFA, appear to have ZERO market share.

When Chris Mellor reported this in The Register  he wrote "What do we notice? Neither Cisco nor Dell appear, meaning their AFA revenues are less than $50m." when of course these vendors were excluded from the list.

If as astute an industry observer as Chris can make this false assumptions customers will belive the competitor salesman when he says it's unsafe to by a Dell or Cisco AFA since no one else is, look they don't even show up on the Gartner numbers.

I implore my friends at Gartner to adopt the duck identification system for AFAs. If it walks like an AFA (uses solid state memory not disks), and quacks like and AFA (has a minimum set of features to qualify as an array like controller redundancy and snapshots) than it is an AFA even if the vendor will let a customer add hard drives later.

The current situation harms the industry as a whole.


- Howard

Re: Gartner All-Flash Array Magic Guadrant: A Bad Influence

Well, this is the first I'm hearing about this, but I couldn't agree more with your perspective, Howard. The idea that analyst reports can have undue influence on the industry or be based on arbitrary criteria (or fudged numbers, which is another story) is obviously a fear that everyone has, but it looks like we have a tried and true case of it here. I can see, way off in the distance, where Gartner is coming from - "All Flash" is, after all, right in the name, and no doubt they're banking on a future where the AFA has a dominant role in the data center - but when we're talking about systems that are so similar, there's just no point in making arbitrary distinctions. The fact that this definition intersects with the business case at no point basically sums up the whole problem here.

When vendors are going out of their way just to skirt your rules, then you know you have it backwards. Restricting what kind of upgrades they're allowed to sell their customers when the hardware is capable ought to raise the biggest of red flags. The only counterpoint that comes to mind here is that it doesn't sound all that hard to skirt the rules. The hybrid version can just run nearly identical software, the vendor can just make one model that's dedicated all-flash (to fit Gartner's definition) and one that's all-flash but can have spinning disks added later, bill it as 'hybrid' and still be within Gartner's rules, no? Not such a big deal - but still pointless, and adding needless complication for customers. Let's hope Gartner is listening.

Re: Gartner All-Flash Array Magic Guadrant: A Bad Influence

Apart from the fact they do add features to address like deduplication and compression to maximize capacity utilization. I am not sure if All-Flash array can really overcome data placement challenge, where administrators can place data depending on its performance demands.

Overblown and Disingenuous Rant!

Firstly the GMQ for SSAs is not the only storage related GMQ that they publish and it probably won't be the last due to industry dynamics.

Secondly they are very much point-in-time based on the previous years activity using a methodology that measures a completeness of vision and ability to execute, natuarally the  incumbents will have a strong advantage for obvious reasons, however Gartner have taken to issuing an acompanying 'Critical Capabilities' report highlighting the features and functionality that some vendor products may have over others, the GMQ should not be read in isolation.

Lastly, I think there is a clear distinction between engineering at the flash chip level and populating shelves with commodity SSDs from Intel or whoever.

Suffice to say Gartner also produce an MQ for 'General-purpose disk storage systems' that are designed to address shared storage requirements within a single system that can support multiple virtual or physical servers, hosted virtual desktop (HVD) infrastructures, multiple databases and files, which includes midrange, high-end and network-attached storage (NAS) systems and hybrid arrays, where many of the vendors which you highlight show up.

Furthermore, I don't see any vendor or prospective purchaser being unduly influenced by any GMQ as a lot of today's purchasing is still largely governed by cost.

I'm also not sure where this rant eminated from, the inaugural GMQ for SSAs was first published 8/28/14 (based on 2013 figures), the next one is not due until 8/28/15, have you been made privvy to this?

From my observation this so-called market (if you can call it a market) was driven by Pure Storage and the GMQ in this respect is merely a reflection, not the driver.

Great points

You're right, Howard.  The all-flash array Gartner Magic Quadrant does do a disservice to customers -- more flexible solutions now have a disincentive to market their capabilities simply because someone at Gartner has decided on an arbitrary definition.


-- Chuck


Re: Overblown and Disingenuous Rant!

Mr. Return,

I'm perfectly happy with the MQ approach. My problem is that there's no logic to Gartners distinction between a SSA, as they insist on calling it, and an array that happens to be all solid state.  So when they publish quarterly sales for SSAs people read stories that say Dell didn't sell enough all flash systems to even be counted when in reality those systems, which are every bit as all flash, engineered and qualified to be SSAs as the 3Par 7450 except that they don't have a seperate model name. 

When people read that some of them take Dell off their list. It's a diservice to Dell and more significantly to the customer who was mislead. It's especially a disservice because Gartner reports the all flash numbers in their general purpose array reporting. Why not issue one press release with total AFA sales not seperated by AFAs that are a different model number from those that are a configuration that doesn't have a seperate model number. Customers don't care if it's an FAS with all SSDs or a VNX-F with all SSDs. Those are directly competing products and if you define a market so one is in and one is out you're not serving the customer.

It was in fact those news stories that reminded me I hadn't written about this yet explaining the timing you find so interesting.

I agree there's a big differnece between building your own flash controller and modules ala Violin or IBM and buying off the shelf SSDs but that's not a factor in the Gartner distinction. Pure buys SSDs as do EMC, NetApp and Dell. Pure's systems are in, Dell's are out, some of NetApps were in and some were out. Now that the all flash FAS is a different product name they'll be in.

As I said in the piece several storage vendor execs have told me that Gartner's rule that an SSA (to use their term) must be a seperate product line has affected which products they bring to market. It means a customer who buys a 3Par 7450, instead of a 7440 with all SSDs is screwed if they want to add spinning disks later because HP bowed to Gartner and won't allow it.

As analysts we have to serve the customer. The "no spinning disks ever" line Mr. Unsworth drew in the sand is anti-customer. 

By the way I'll take overblown but disingenuous? Did I really pretend to know less than I do? That would be a first.

Thanks for reading

- Howard


Well said, Howard

I've been watching Gartner for well over 20 years, nad in all my time, I've never been able to understand why vendors bow to them, much less why customers pay a bloated premium for their advice. I work as a Competitive Intelligence professional, and I can categorically state the Gartner gets enough wrong with every vendor they rank to be concerning.  

Pile on top of that your very good analysis of how an MQ criteria is constructed, and it looks as if whomever spends the most on lunch for them gets to define the terms!

Thanks again for wrinting this. I'm sure Gartner is not amused, but this is exactly what is needed to spur vendors and custmers alike to reevaluate their analyst spend.

Good Call

Excellent write up Howard! I also see that Nimble is making some waves about the issues with these types of reports as well.