Hyper converged is just the tip, let me show you my rack

To say that the Hyper Converged appliance marketplace is “so hot” right now is more or less understatement of the year. Nutanix, Simplivity, Scale Computing, Nimboxx, Pivot3, EVO:Rail, and now Stratoscale (who just secured a 32M B round without even having a reference customer that I can find) are all vying for a piece of a very large pie.


Estimates place the Hyper Converged market at around 10-20 Billion (with a B) of TAM (Total Addressable Market), and frankly that may be just thejustthetip tip.  Given the amount of technology that can be condensed (and displaced) into a small appliance form factor, those data center dollars start to add up awfully fast. Reducing the annual spend on multiple point solution based appliances with a small data center building block that does all of them is appealing, and can significant reduce TCO. Taken together, the disruption will continue as more organizations adopt a more building block approach to a good portion of their virtualized workloads and it makes sense that Hyper Converged offerings will be looked at as the key platform of adoption. Hyper Converged players are packaging the capabilities of multiple appliances into a single box. Storage Array, Storage Networks, Servers, Backup appliances, Cloud Gateways, and WAN accelerators are the main focal point for takeout, but also replication and backup software can be targeted as well depending on how robust the platform is in terms of convergence. All told, the standard Hyper Converged system will replace around 4-8 appliances and/or software systems to consolidate into a single appliance based model that scales resources pools in a predictable manner. This is why in a few short years we have seen a space that didn’t exist, ramp up to be one of the hottest technologies available. But buzz and press are one thing, are customers buying into this model at the pace being claimed?

Don’t Believe The Hype

dontSo lets slow down a bit and step back from all the hype. Hyper Convergence makes sense for a lot of organizations. It certainly has a natural appeal to the virtualized data center, and I see a very large install base of potential customers that would look to leverage Hyper Converged systems as their main “unit of infrastructure” (ooh I should trade mark that one)  Other use cases for larger Enterprise customers are ROBO/DR targets and Dev/Test. The prime targets for a Hyper Converged play revolve around IT orgs where there are a handful of people wearing all the hats, or the IT Generalist crowd, where infrastructure decisions are made by small groups that can control the decision making process.  Alas, this is not where the bulk of IT dollar spend exists today. While that space has the sheer numbers of customers, it cannot begin to comprehend the amount of annual IT spend that the big players go through in a year.  There is a group of around 200 or so global companies who have annual IT Budgets in excess of 1 Billion, for which the Hyper Converged systems would simply be looked upon as toys.

With Nutanix claiming 200M in sales it’s easy to think that they have ‘arrived’,  but when you see individual deals for specific infrastructure projects or deals that are twice that size for a single customer, it’s hard to buy in entirely that the data center of the future will be of the Hyper Converged model. In that space risk aversion rules the day, and the lack of maturity for Hyper Converged systems is not there yet.

Let Me Show You My Rack

I’ve purposely left off bare metal workloads, and how Rack Scale addresses them and how Hyper Converged cannot, alas that is a post for another day. So while there is a great benefit to Hyper Convergence, there are alternatives at play that address those 200 customers where the smaller appliance based systems cannot play, and now we are starting to see  an emergence of Rack Scale based infrastructure offerings. m-seriesI think that many people would place vBlock and Flexpod into this space, but in my mind those are still multi-vendor reference architectures / rack architectures that still rely too heavily on multiple vendors to integrate, design and deploy. The coming wave in rack scale architecture will come in a form factor that deconstructs the traditional multiple appliance based infrastructure approach with one that is more modular in its nature. I recalled reading the “Rack Endgame”  by Stephen Foskett a while back as well as the follow on article Cisco’s Trojan Horse,  both illustrate some of the potential that Rack Scale architecture can achieve, and both offer a first step forward to understanding where Rack based convergence is moving.  Looking at UCS M-Series Modular servers as well as HP’s Moonshot programs as the first two major players to embrace this space and bring products to market. Both are looking to productize the concepts that were being developed by the Open Compute project and Intel’s Rack Scale Architecture approaches. Not to be left out, I see EMC making a full effort to enter this space and address their current favorite customer that wishes to utilize Hybrid Cloud in their organization. This is where the EVO:Rack based systems will come into play for not only EMC, but other large vendors as well if they choose to build systems around that specific architecture.

Given that this entire post materialized in about 10 minutes from reading that Stratoscale article, and I felt a deep seated need to get all these thoughts out of my brain, I’ll leave todays post at that. As always, there will be more to come.