Finally. Finally. I’ve been waiting a while to put this post together.
Most of you have read about the dustup between Tesla founder Elon Musk and the NYT reporter who wrote a critical review of the car’s performance. What was interesting to me was that the conversation was immediately framed with an incredible amount of empirical data, courtesy of the sensors in the car. Tesla was able to clearly show data that didn’t correlate with the reporting, and one would think that would be the end of the story.
But of course, it wasn’t. The reporter, even when faced with the data, refuted the Tesla characterization of his story. What we learned was that while the data can tell us what someone did, it can’t tell us why they did it. The data can’t ascribe motive. Was the reporter a douchebag bent on writing a negative review? Was he really lost in the dark of a rest stop? Was he really trying to behave how “normal” owners of a $100k electric car would behave? The data, no matter how granular, can’t answer any of those questions, can it?
Much the same drama is played out in the converged infrastructure marketplace, but until recently we didn’t have good data to work from. Everyone was speaking to whatever narrative their marketing team made up, which lead to lots of confusion, FUD and delayed adoption on the part of customers. Not a good place to be, for anyone involved. In late November, Gartner published the first ever market share report for integrated systems. Finally. Rather than making assumptions about the size of the market, or the success of the players involved, we can go to a “trusted” source and use them to help define the landscape, at least in terms of revenue over a period of time. The report, if you haven’t seen it, is freely available (No registration, and no pay wall, which is how it should be…) on the VCE website. A big thank you to Bob Wambach and our Gartner team for making that happen!
So let’s dig into the data. First, Gartner has done what we’ve been trying to do for two years: provide three clear market segments. First, there is the “Integrated Infrastructure System” which is a fully integrated, productized solution that includes server, storage and network components. The players in this space include VCE, IBM, HP and Dell. The second segment is the “Integrated Application/Workload System” which are appliances pre-integrated with a specific database or application, such as those from Oracle, HP and IBM. Finally, we have the “Integrated Reference Architecture” as the third segment, with only the FlexPod being called out as a provider.
Interestingly, Gartner spends two full paragraphs defining the reference architecture segment. First, they separate the specific configurations called out by the RA vendors as being more detailed and specific than the general hardware compatibility testing done by the individual vendors, which is a reasonable statement. They also make the case that the Reference Architecture products are in part defined by being positioned against the Integrated Systems and Integrated Workloads offerings. This is also reasonable and supported by the activity we see in the market. Gartner makes the case that these are still different offerings, delivered in a different way to different customers, but that they’ve included the Reference Architecture segment for readers to include or exclude as their perspective warrants. Personally, I’m thrilled they are included because it helps rationalize the segmentation of the market, and the Reference Architectures are certainly a choice customers have.
First, let’s look at overall revenue generated over the six consecutive quarters that Gartner covered. Infrastructure Systems (IS) is the largest of the three segments, with a slight lead (3%) on Workload Systems (WS) and a more sizable lead (11%) over the Reference Architectures (RA). The good news is that all of the segments are growing faster than the global infrastructure market, showing customers moving to a converged model in significant numbers.
It’s also important to note that while revenue in 2011 came in at $2.9 billion, it was still a very, very small (3.5%) part of the $83 billion overall market. The message there is that every vendor that has entered the ring has a sizable amount of runway before saturation in the market leads to more head-to-head competition.
When we dig into each section, we get a good idea of the overall direction. In the IS segment, VCE is the clear leader in market share (57%) and was the only vendor to generate over $1B in revenue in the six quarters that Gartner covered. We can also see that VCE continued to increase it’s market share over the course of the study, starting at 53% and gaining 4% in the next 18 months. HP (-2.5%) and Hitachi (-7.2%) both lost market share, while Dell (1.4%) and Oracle (0.7%) both made headway. IBM didn’t introduce PureSystems until 2Q12, so there’s no way to determine how much they have grown their 2.8% of the market until the next report comes out. We also see that HP and VCE stayed at or above the overall growth rate of the segment, despite the differences in actual revenue generated. This market will be interesting to watch, as we start to see more numbers from IBM in particular. Gartner needs to be careful how each of these vendors gets classified long term, because with customers being able to, and encouraged to, use alternate or existing hardware, the line between integrated system and reference architecture becomes blurred. Frankly, since all of the vendors here allow customers to swap components in places where they aren’t market leading, and even when they offer a fully integrated solution, it’s hard to figure out which is which. Hitachi for instance, offers their Hitachi Unified Compute Platform in a “Pro” (integrated) and “Select” (reference) version, much the same way HP, IBM and Dell does, so it’ll be difficult to pull out revenue statistics from each flavor. Those companies could end up doing a lot of business in the reference architecture segment, depending on how the revenue is reported. Oracle is the other interesting case here, since nothing they manufacture on the hardware side is capable of general purpose computing workloads; everything is designed for Oracle apps. How is that not part of the Integrated Workload segment? Maybe we’ll see that shift over time as well.
Over in the WS segment, the vendors who have the ability to control both the hardware and the software have a clear lead. Oracle in particular, has been aggressive in providing appliances for it’s own applications, and their success is evident.
What will be interesting is to see how this shifts in 2013. With the proliferation of SAP HANA appliances overall and with the introduction of a specialized systems line of Vblocks, there may be increased competition in this segment. Unlike the other segments, the “Other” category here is significant, ranking second in total revenue, so there’s a good chance that additional players could emerge.
Finally, over in the RA segment, there’s FlexPod….and everyone else. Cisco and NetApp were the first to recognize the need for this kind of offering, and have really become the standard for this kind of deployment. FlexPod owns 79.7% of the RA market, although the relatively sluggish 37.4% growth rate could signal coming change. EMC has gotten much more aggressive with it’s VSPEX offering, and as FlexPod has moved into the lower-cost configurations Gartner notes that their average sale price has decreased. Both EMC and FlexPod like to tout the number of customers they have in their RA programs, but with very loose guidelines around what constitutes a “certified” customer it’s hard to tell what reality is. For example, if I have a NetApp storage array and UCS blades, I’m not technically a FlexPod customer, but if I add a Nexus switch upstream I suddenly am. Does all of that revenue count as a FlexPod? VMware isn’t a requirement for a FlexPod, but if I include VMware licenses with the purchase from my channel partner, does that count as FlexPod revenue? If a customer bought their Cisco gear from one partner and their NetApp gear from another, does the spend with both partners count as FlexPod revenue? What if the second partner in that case wasn’t certified to sell Cisco DC products, does that matter? With no real definition it becomes hard to understand exactly what is being referenced. But even if we trust the Gartner numbers here implicitly, the revenue per customer in the reference architecture segment is obviously much, much lower than in the other segments.
Overall, the data speaks volumes, and can inform us about the past, but doesn’t do a great job of predicting the future. It doesn’t keep idiot marketing folks from inventing context that doesn’t exist. It doesn’t ascribe motive. But it does, finally, start to put a framework in place that we can use to measure going forward. And that’s something. Kudos to Gartner for the report, and for letting VCE make it available to the public!