2014: Validation of Hyper Convergence – 2015 Hyper Converged Goes Mainstream

2014: Validation of Hyper Convergence

Screen Shot 2014-11-27 at 12.38.37 PMLooking back on 2014 I saw it as  the year that Hyper Convergence  has come into its own as a legitimate data center technology. It’s rare that I don’t run into a customer in my day to day discussions that is not evaluating one of the current Hyper Converged vendors, or looking to adopt or develop a strategy for how Hyper Converged can become part of their IT infrastructure to some extent. I chalk this up to three main factors that continue to be the major challenges that IT faces today :Complexity – Data Growth – Scalability. The Hyper Converged systems today aim to address those three areas, but they also are seeking to become the standard building block unit for most virtualized workloads.

Delivering Simplicity

simplicityThe Hyper Convergence promise is simplicity. That’s the first and foremost value proposition provided to the end user. The ability to plug in, turn on, and deploy workloads in a time frame that surpasses the traditional multi-vendor reference architecture deployment model is a strong motivating factor for customers who are looking at the Hyper Converged appliance based model. This is especially true if they have a strong understanding of their workload needs, and are looking at a fairly static growth pattern. That is one of the key winning aspects of a Hyper Converged solution over the engineered Converged Infrastructure plays like vBlock and Flexpod. Compared to the first generation Converged Infrastructure systems, the design aspects involved with Hyper Converged systems often are a driving factor in its adoption as well. You need 5000 VDI seats? Simple, here is a 400 desktop based appliance, go do the math. To some that may seem as an oversimplification, but that is indeed the approach that many Hyper Converged vendors will take when engaging with customers, and at times it’s hard to argue against that approach. The ability to design around pre-determined chunks of compute and storage, ie: the lego data center approach, is another strong factor in favor of Hyper Convergence. The abstraction of the complexity that goes into making all of the interwoven pieces of the legacy stack work together is removed, thus the removal of complexity of design, this all points to the ability to deliver simplicity.

A Jack of All Trades



The “one size fits all” approach to the data center is not without its pitfalls. Sure Hyper Converged solutions are simple to design, easy to deploy, and  are well suited to a majority of virtualized workloads, but I’d hazard a guess that most data centers are not a simple homogeneous environment, and that is why in the first two years of their arrival, the solutions have been targeting the specific use cases around ROBO, QA/Test/Dev and VDI. Much like the same way that Virtualization made its inroads into nearly every IT shop on the planet. this new way of providing infrastructure is going after the low hanging fruit, yet I see adoption and acceptance occurring at a more rapid pace than what VMware faced from 2004-2009.  One could ask, If the appliance based Hyper Converged solutions address all of our data center needs, then why are we not seeing it become the defacto  infrastructure platform adopted today?

Well, this is where I get to throw out, “it’s complicated”, because in theory Hyper Convergence can address almost any workload that is virtualized today.  Still, the entrants in the space are young, and that means that the focus will first and foremost be on the lower risk spaces of ROBO, Test/Dev, and VDI. This isn’t necessarily a limitation of its capabilities, but more so a realization that risk aversion tends to rule the day when it comes to production infrastructure. I do not expect this to be the case for 2015, especially since VMware is enabling so many OEM groups to become market participants without having to do much in the way of actually creating anything other than the vessel of delivery.

Prediction Time: 2015 Hyper Convergence Goes Mainstream

On the last episode of In Tech We Trust, we discussed Hyper Convergence in detail as part of the 2014 Trends, and I offered some of my thoughts on where I see the space going. To elaborate a bit further, I would expect a few things to happen this year.

nutanixI’d look for Nutanix to IPO this year. Hitting the 5 year mark, with nearly 3 of those with a shipping product bodes well for them, along with their published run-rate of business hitting the 200M mark and a 2 Billion Valuation. Stronger expansion into Europe and APJ, coupled with the reseller agreement with Dell are also beneficial. I’ve also seen a toning down of some of the more caustic rhetoric from some of their team members, that tends to be a sign that the adults have taken a stronger leadership role. All of this points to the move to go public. I personally believe they should have launched their IPO prior to VMware announcing EVO:Rail, but  the case can still be made that the momentum of VMware’s EVO partners going into the market, and a general shift for standard virtualized compute adopting more of the “Software Defined” aspects will offer numerous reasons for the street to see investment value. Does this mean I think that Nutanix is the clear winner in the Hyper Converged space, no it does not, but for now they have the strongest mind-share, and a diverse enough offering to give them top billing.

simplivitySimpliVity should be looking towards a very large D round of funding that will facilitate further growth.While they do not get the kind of marketing and community buzz that Nutanix does, SimpliVity has a very robust platform and significant intellectual property in their portfolio. Their partnering with Cisco is a much smarter move in my view than the Nutanix/Dell relationship (more on that on another post), and will allow them to leverage the large UCS installation base and strong Cisco partner network. What remains to be seen is how much integration between the two platforms will be done as there is no direct UCS integration right now, though I believe it would be prudent for them to work on that quickly. The original platform based on Dell worked to launch the product, but I see the Cisco relationship being the one that brings them a larger measure of success and awareness. I’d also look for them to announce a secondary hypervisor support in 2015, most likely being Hyper-V which has good traction in the SLED/FED space (an area I see SimpliVity having much success in).

evoEVO:Rail will be great for VMware, but not so great for the OEM’s that have latched onto it. As a 1.0 product that has the added benefit of being able to take a wait and see approach to the Hyper Converged market, I was a bit disappointed in what eventually will hit the street for customers. EVO:Rail is going to be a very expensive solution, even at the lower end of the OEM spectrum with manufacturers like Super Micro being part of the group. While not all the OEM’s have pricing out, I’ve seen 180k as a starting point, which puts it as one of the more pricey solutions based on the number of VM’s or Virutal Desktops that are currently being supported.. Yet even as I say this, it will be hard to beat the marketing muscle and ability to penetrate into the customer space that VMware can enact when coupled with the major OEM partners. Once again, VMware has dictated direction to the OEM’s, this time the server vendors. In the past it was directed at the storage companies with VAAI, but now, there is greater profitability to push more of the “Software Defined” Data Center into customers hands and to have a dominant position as the paid Hypervisor of choice. EVO:Rail as its constituted today is not a very solid Hyper Converged platform, but the fact that HP, Dell, HDS, SM, and others now have a “good enough” entrant to compete with Nutanix and SimpliVity (even at the cost of their own home grown/partner solutions) is a big enough deal to make it the solution that SimpliVity and Nutanix will have to compete against as well as the traditional status quo of legacy rack solutions. I plan to expand on this in greater depth with its own blog post soon.

federation-logo-largeThat leaves the great unknown being the EMC Federation, and what comes from EMC/VCE for their entry into the Hyper Converged space. I think its easy to see that they will use EVO:Rail as one vehicle, but I’d also look for some kind of new solution that would look to leverage the ScaleIO technology. I’d go one further to say that their larger focus will be on the RackScale technologies.

Closing Thoughts

For a majority of virtualized workloads, and for customers looking to heavily virtualize,Hyper Convergence is a good fit. The “simplicity” message resonates strongly, as does the TCO argument which can in some cases be 3X reduction across the board. True there are caveats about maturity of the vendors involved, but if we see more VC money pour into the established, and up and coming players, along with an IPO by one of the founding groups many of those fears will subside. Most notably though, the validation of the Hyper Converged approach by VMware entering the market with EVO:Rail, and the rapid adoption and partnering by the major server OEM’s, points to this being a true disruptive technology transition that is here to stay.

Cloud dynamics are going to play a factor in this market as well, as we see companies adopt a Hybrid Cloud approach or move into Public First or adopt a Full Private approach. Yet that battle space is still being identified, and looks to be more relevant to a specific subset of very large Enterprise customers.

2015 will be the year that many companies start to look or at least entertain a “Hyper Converged First” approach to specific workloads, and use cases. What remains to be seen is what impact RackScale and Hybrid Cloud play in the designs (that will be a new set of blog posts all to themselves). And as a last note, the changes with technologies such as Open Stack, Docker, etc. will play a big part in how our next generation data centers are designed, and that is where I can see a robust “cloud ready” appliance based on the Hyper Converged model or Rack Scale approach become the defacto standard for infrastructure deployments.

I’ll have a lot more to say about this space as the year goes on. I’m working on several deeper dives into the Rack Scale side of things and hope to be able to provide some content soon. Thanks for reading.

7 thoughts on “2014: Validation of Hyper Convergence – 2015 Hyper Converged Goes Mainstream

  1. Stuart Miniman
    January 5, 2015 at 6:05 am

    Hi Gabe – good commentary. As with any hot technology space, the competition is really starting to heat up. There are more startups that you didn’t mention (Nimboxx, Gridstore, Scale Computing, Maxta) and more coming in 2015. The trend is on the radar of all of the big players – HP has made some moves and every player will have to have a partnership or answer for users ASAP or they could lose some business. I would clarify that Nutanix has an OEM relationship with Dell and SimpliVity is the driver of the Cisco relationship; it will be interesting to see which is more fruitful – Dell has not shown the ability to sell converged yet while Cisco is the leader in converged with partnerships with every storage vendor, we’ll see if this translates to hyperconverged with SimpliVity, Maxta, VSAN, etc.
    2015 should be the year that hyperconverged solutions get a lot of traction, as you point out, how it fits into various hybrid cloud options remains to be seen.
    Exciting times, I look forward to the discussions.

  2. duncan@yellow-bricks
    January 5, 2015 at 7:57 am

    Two things:
    1) not sure where your pricing comes from but I have seen different numbers. Also, not that this price includes all VMware licenses and 3 years of support for the 4 hosts. (8 x vSphere E+, LogInsight, vCenter, VSAN etc)
    2) Number of VMs supported… There is no number stated for EVO:RAIL. There is a number of what the team has tested with and could comfortably run, but this will of course depend on your workload. If your workload is different then results will be different as well.

    1. January 17, 2015 at 11:58 am

      The dark horse is Pluribusnetworks.com with its hyper – converged infrastructure solution that fuses network, compute, and storage, leveraging Openstack and a allows folks to config the network like a server.

  3. January 5, 2015 at 8:09 am

    @Duncan: thanks for the clarifying points, I know that support is included, that certainly having the support be part of the discussion is important.

    As for pricing, I’ve seen limited data online. I would have expected the pricing to be announced by now, but it seems like a game of chicken is occurring when it comes to public announcements. I know Howard Marks had written a few things about pricing as well.

    As for the number of VM’s I think that point is important to expand upon. 192GB of RAM is not significant and is the limiting factor as I see it with the EVO design. 256-384 would have helped with density and given a bit of wiggle room, or at least having the option to expand the RAM available. 25 VM’s per host isn’t a bad consolidation ratio, but taken in the greater cost context and based on what a customer could buy with a non Hyper Converged solution it certainly becomes a factor.

  4. January 6, 2015 at 10:07 pm

    Hi Gabriel

    Very good post, indeed. We work mainly with SMB customers and i see more and more acceptance for Hyperconverged solutions in this area. SMB customers can’t cope with the complexitiy of non-coverged systems any longer.

    Looking forward to your next post.


  5. January 26, 2015 at 2:21 pm

    To my knowledge, ScaleIO is going be build into the esxi kernel, just the way vsan is. Vsan is more like a SMB solution compared to ScaleIO, supporting 32 nodes, where ScaleIO do 1000 nodes.

    2015 is certainly going to be an interesting year 🙂


Leave a Reply