Today Cisco announced that they are launching a new hyper-converged offering (Cisco “HyperFlex”) that is built off their UCS C-series rack mount servers, and powered by an OEM of software from a small startup named SpringPath.
I to give want quick congratulations to Cisco to re-entering this market, and to SpringPath, a huge step and big deal for a relatively unknown company and IP stack.
A couple quick thoughts – in usual verbose form :-) - as Virtual Geek readers internalize and reflect on the news:
- It’s an indicator of what we believe, and reinforces what we are doing. We believe that customers increasingly won’t build infrastructure – they will consume converged and hyper-converged stacks. Furthermore, we believe that ultimately even converged and hyper-converged infrastructure should be invisible layers behind turn key IaaS/PaaS/Data Fabric offers. This means in the long term, if you are in the infrastructure business, you HAVE to be in the a converged/hyper-converged business – so Cisco has to do something. In addition – while we think that Converged infrastructure is important (and will continue to be the majority of the market for a while), we believe that software defined, hyper-converged offerings are a growing part (starting smaller than CI, but growing faster) of the infrastructure ecosystem. Let there be no doubt: EMC aims to be the leader in hyper-converged infrastructure with VxRail and VxRack, just like today we are the undisputed leader in Converged Infrastructure with VCE Vblock.
- It’s funny that finally people are waking up to the fact that the “unicorns” are relatively small still, and many have business models that tend to operate at a material loss.
- Vblock has quarterly revenue growth that dwarfs large parts of what the entire the startup ecosystem does in revenue in a full year – and does it profitably. I want to make something super-clear – because customers (and ultimately business) talks. When EMC publicly stated a $2B+ VCE run rate in January 2015 – that translates to roughly $500M (+!) of business in Q4 2014 – simple math, divide by 4. Likewise, when EMC publicly stated at $3B VCE run rate in January 2016 – that translates to roughly $750M (+!) revenue in Q4 2015. That’s a huge growth rate for our Converged Infrastructure business year over year – where else in infrastructure do you see that sort of growth and absolute numbers (other than public cloud)?
- …And that was all before we got started in the Hyper-Converged space!!! While Hyper-Converged systems are a newer domain for us, when it comes to Hyper-Converged, customers that need huge scale (hundreds to thousands of nodes) are choosing VxRack. Customers at smaller scales (starting at <$60K list for a 4-node cluster!) and scaling up to tens of appliances (64 nodes!) are choosing VxRail – I think momentum is in our favor. I don’t say this lightly, and I know we cannot be arrogant or rest on our laurels where we are strong, but instead must continually disrupt ourselves. I have a great deal of respect for Cisco, and I have a great deal of respect for all the Hyper-Converged startups. In particular, I want to give a shout out to the Hyper-Converged startups - they forged the way in the early days – and I respect them a great deal, but now it’s on like Donkey Kong. Competition is good.
- Yes, you’ll hear a lot of hyperbole from the press and analysts, but we won’t over-rotate and over-react. When Cisco made their Whiptail announcement over a year ago (which wasn’t a simple OEM, but an outright acquisition!) many people thought that would be the end of the Cisco/EMC relationship, and certainly of VCE. Instead of reacting emotionally, and over-reacting when the Whiptail thing happened – we did the opposite. We shrugged our shoulders instead of freaking out – doubled down on execution and working with Cisco, and a couple years later, our VCE business is 2x larger (and Cisco’s business with EMC tracks with VCE – and today we are a huge component of Cisco’s UCS and Nexus business). … And Whiptail is, well, everyone knows how that story ended.
- I understand that Cisco wants to have a presence in the rapidly growing hyper-converged space, and it appears their other hyper-converged OEMs weren’t answer enough. I have no idea what this means to Cisco’s other hyper-converged partners. Cisco needs to have an offer in the hyper-converged appliance space, a route to market. Can VCE be that route to market for Cisco products? Absolutely yes for Nexus, not right now for Cisco UCS C-Series. Yet Cisco UCS is sacrosanct to us as the compute layer in Vblock. Question: Why? Answer: because in scale-up converged infrastructure, customers like B-series architecutrally, UCSM, and the tight integration with the network domain. The customers tell us so – and we listen. Likewise we use Nexus switches in Vblock, VxRack and VCE VScale Fabric. Why? customers like it, and tell us so. However, we didn’t select UCS C-Series rack mount servers in VxRack and VxRail based on our point of view that in these hyper-converged systems are use cases where low entry cost, broad configurations rule the day. These systems are all about the software stack – not the hardware layer. Things that make sense in large scale converged systems (like tight hardware coupling between servers and the physical network and linking the server/network management model) makes sense in UCS B-Series – which is why the two go together so well. Is the same true in hyper-converged models? My personal point of view is that those same ideas simply make far, far less sense in hyper-converged models. Why? In that domain, the hardware itself is very focused on “industry standard” and “commodity, but wicked awesome supply chain” – and the value is in the software stack, and how the whole offer comes together. I still think the above is all still true. Regardless of whether we didn’t select UCS C-Series for VxRack and VxRail - of course Cisco must have an offer here, so I get them deciding to go this way.
- I think that Cisco’s strategic view point on the topic of infrastructure convergence, is different than ours (and that’s OK). Cisco’s strategic point of view seems to be one that convergence can ultimately work in an “open ecosystem” models. This means they partner widely - with EMC more than anyone, but also with NetApp, with IBM, with Simplivity, with Pure, with SpringPath heck pretty well anyone. I suppose that partnering broadly makes sense if you don’t think of convergence as “bundling” or a “route to market for products”. We don’t view convergence that way. We view convergence as a fundamentally different way of consuming (buying/designing/supporting/sustaining/engineering) infrastructure. EMC’s point of view is that convergence and hyper-convergence is not a “side show” or a simple “route to market” – we think it’s the “plot” of what’s happening in infrastructure, along with industrialized/curated IaaS. We think that converged/hyper-converged infrastructure (that including new composable hardware models which is a new variation on the system design, but fundamentally the same IDEA) will be the form that MOST infrastructure is consumed ultimately. Furthermore, we believe that when a customer goes converged – it means they are looking for something different – they want answers, not choices. They want to consume, not build. It means things need to be engineered, manufactured, supported, sustained, and managed as ONE. That means that “ecosystem” models work ON TOP of CI/HCI – but not so much INSIDE CI/HCI. Like it or not (and you may not, and would love to hear thoughts/dialog!) – if you believe our strategic point of view, this means that convergence runs counter to the idea of “partner with everyone ecosystem models”. Don’t get me wrong - EMC embraces choice. We partner in open ecosystems, and always will. We just think that model is the “open systems, you build it, you sustain it” model that has been with us for 20 years. We think that when a customer wants a turnkey converged stack, we think you cannot build a strategy of “turnkey delivery” built on “whatever – you tell us what you want”.
- We’re excited by the announcement – seriously. IMO, the people who will struggle the most with the mega announcement of VxRail and today’s announcement of Cisco’s entry in the market is not EMC. Today’s news seems to make life harder at the startups/”unicorns” whose opportunity window is closing. Their choices are narrowing, their business is going to be more challenged than it was yesterday, and their strategic exits are getting harder. It’s not just me – lots of ink on the topic in the last few weeks – that I won’t link to as it’s just piling on. People are starting to wake up that many (most?) infrastructure startups don’t have sustainable business models, will need to adapt, and even VC money doesn’t grow on trees. Some of the pervasive speculation of possible exit scenarios for some seem less likely today than they were yesterday. I’ll say this again – I do not underestimate the startups, they are pushing, innovating, and can (for some time) operate with material losses – which means they often appear much larger than they are. I also don’t know the future – but I do know I wake up, and everyone at EMC wakes up every day to drive towards infrastructure leadership on all fronts. We’re far from perfect, but we’re working hard. I hope the startups are arrogant and underestimate us :-)
My guidance to EMCers and EMC partners is the following:
- Be cool as a cucumber. Relax!
- Reach out to your Cisco colleagues and give them a hug. They will be very excited with this news, particularly the Datacenter specialist team – at first. They are our brothers/sisters – so their reaction is no different than ours (or any ecosystem) when a new announcement comes out – excitement! But being in the converged/hyper-converged business is not easy, and it’s a marathon, not a sprint. Our relationship with Cisco is strong and most importantly is extremely relevant in the market and with our mutual customers. Trey Layton (VCE CTO) wrote a great blog today as well I would encourage you to read here. EMC and Cisco enjoy great success together, and that's not just coming from EMC. If you haven’t already seen it, check out the recent blog from Frank Palumbo, Cisco SVP DataCenter Sales: “Cisco and VCE Continue to Lead the Industry”. We love partnering with Cisco. Why? Pretty simple. When we partner, we add more value for our customers and that is what our customers want.
- Our partnership with Cisco continues to grow – Compute and complete network in Vblock, physical Network in VxRack.
But – I want to be clear we are opinionated and feel strong about our offers in the Hyper-Converged space. It would be disingenuous to suggest that in some cases Cisco and EMC will compete in this arena. For perspective you should visualize a scale with $3B with a 30% CAGR on one side, and a small feather on the other side of the scale)
When it comes to our offers in the Hyper-Converged market VxRail and VxRack are strong, competitive, innnovative offers:
- VxRail – simple, starts small, and is the easiest way to deploy our customers’ standard (VMware) at the Enterprise Edge and SMB/SME targets. Yes, some customers want heterogeneity – but generally not when they are small or at the edge, where they want simple, rather at the enterprise core. I’m not an expert on SpringPath (didn’t know about them until recently) – so the below isn’t about them. When it comes to the hyper-converged appliance domain, I think VxRail will:
- Perform really well – you just don’t want certain parts of transactional IO stacks in userspace/guest worlds (happy to keep giving independent reviewers VxRail systems)
- Start smaller (at far lower entry price points and on the whole scaling curve, but also on environmentals like space, power, cooling)
- Scale better than most of the hyper-converged players (64 nodes in a cluster)
- Be able to mix nodes in clusters better (any mix and match configs today – and VxRail is going to get a broad range of form factors beyond 2U4N quickly)
- Integrate with VMware more than anyone – after all it’s a single combined engineering team.
- Support the all-flash world really well – and these days, this is the sweet spot for transactional infrastructure. Storage stacks (software defined or hardware) that optimize for hybrid vs. those that optimize for all-flash are increasingly going to look anachronistic.
- Replicate really, really well (sync, async consistency groups, WAN optimization, you name it)
- Integrate with cloud object storage – and I think the ability to add capacity without physical footprint is important in hyper-converged offers.
- be supported as a singular appliance better (this is huge – there is no other offer that has single vendor support, for EVERYTHING in the stack, inclusive of the hypervisor layer)
…I want to reiterate that in this list I’m not talking about HyperFlex, I’m talking about VxRail’s strengths. That list of strengths applies to all the hyper-converged choices customers now have – including the new players and the people who have been at it for some time. Only customers, the market and time will tell (doesn’t matter what I or anyone else says – customers talk).
- VxRack – scales big, and scales easily, and is designed for Enterprise Datacenter scale
- VxRack integrates a fully engineered/sustained/maintained network domain. I want to be clear on this:
- In Appliances like VxRail (and unlike at rack-scale), “Bring any network you want” is a feature.
- Conversely, I think integrating the network domain is very important as you SCALE BIG (not 3, or 5, or 8 nodes in a cluster – but 100’s or 1000’s). When I say “the network is important at scale” – I’m talking about:
- The physical layer performance and scaling (which is why we’re pumped about the new Nexus 9200 in VxRack)
- … more importantly, how it ties into the SDN and management domain.
- Remember that when you see “x” in a EMC/VCE product (VxBlock, VxRack) – it means we can integrate and single support NSX. Hint: VxRail’s big brother is coming soon.
- VxRack is OPEN. VxRack is not only completely open in the network domain. It’s OPEN in every dimension:
- VxRack support Hyper-V, KVM, vSphere but also the new world of container-oriented abstraction/automation models (Mesos, Photon Platform, bare-metal) you name it.
- VxRack will support the whole data fabric ecosystem (MongoDB, Map Reduce, Spark, Cassandra, you name it). I think that going forward, people buying traditional proprietary RDBMS appliances for new data fabrics is going to become an intelligence test.
- VxRack has a completely open northbound API and management plane (RackHD – github repo here).
- VxRack supports a huge set of workloads, and supports all sorts of heterogeneous use cases.
- VxRack supports traditional “infrastructure dependent” workloads, and also the new domain of container-oriented “cloud native” workloads.
- VxRack will support all sorts of rack-scale elements like DSSD that are unmatched in the industry.
- VxRack will continue to evolve fast.
- VxRack integrates a fully engineered/sustained/maintained network domain. I want to be clear on this:
We will continue to double down in the hyper-converged market, and with the FTC approving the Dell merger last week, and the EU approving it this week – you can imagine all sorts of ways we would accelerate this space.
We don’t think of “Hyper-converged” as a “product”, we think if it as a “form of system architecture”. We think it will show up in many ways (software-only things like VSAN and ScaleIO, and in systems like VxRack and VxRail for example).
While we will see many 100,000’s of hyper-converged software customers, and many 10,000’s of VxRail customers (SMB/SME/Enterprise edge is a huge market) – what we’re seeing with VxRack is that they are smaller in number (100’s – 1000’s ultimately), but they are HUGE in scale (think hundreds of nodes at a time). Again, only customers, the market and time will tell (doesn’t matter what I or anyone else says – customers talk).
A modern data center must be based on modern infrastructure which is our converged offerings (Vblock, VxBlock) and hyper-converged offerings (VxRack and VxRail). Customers want a set of solutions that are all related, and want a partner who can bring them to life. I really ask – who else can do this:
- EMC has a converged/hyper-converged portfolio in all forms: Blocks, Racks, Appliances (including everything that we do).
- We just don’t believe in “one architecture for everything”. It’s never worked, and there’s no sign that’s ever going to work, regardless of how single-product companies feel or posture.
- Our portfolio therefore needs to be a composite of different architectures, but it cannot be a mish-mash of all sorts of different partnership models, different ecosystems. Even on Vblock and VxRack with Cisco – it’s important to understand it’s not a “meet in market” reference architecture model (aka FlexPod/VersaStack), it’s a OEM model. We are one company, a trusted company, who isn’t in this business “part time” - we think our future depends on it, and we are investing to win. OEM’ing the core IP seems strange to me (it’s what makes a hyper-converged thing a thing), but OEM’ing components can work.
- EMC has the ability to connect Blocks, Racks and Appliances via the VCE Vscale Fabric and VCE Vision – the CI/HCI strategy isn’t an island, because most customers need some of each. I think we as a VCE team have more to do here, and I’m prioritizing and directing resources there.
- EMC has the ability to bring CI to life. Question: What the heck does that mean? Answer: because no one wants infrastructure for it’s own sake. They want it to run app stacks. EMC can industrialize and curate the IaaS, PaaS, and Data Fabric stacks that customers want to use. The VMware vSphere SDDC-powered Enterprise Hybrid Cloud is only the first “curated stack”. I think frankly we can do more to simplify and harden it (and we will). We won’t stop there - there are more to come (VMware Photon Platform, Mesos, OpenStack/KVM, Pivotal PCF, Pivotal Data Suite, HortonWorks, Cloudera, Splunk, MongoDB, Cassandra and many, many others…. Cranking fast J )
Again congrats to the Cisco and SpringPath teams – looking forward to seeing you in the market!
Would love to hear what YOU think – where am I wrong, and what can we (EMC) do better? I’m sure that some other industry players will take exception to some of my statements above – and I encourage the debate dialog – I love to learn!
*Disclaimer - SimpliVity employee
Good post, as usual, Chad. Always appreciate your take (and enthusiasm!)
Quick note and exception to "it appears their other hyper-converged OEMs weren’t answer enough". I assume you're primarily referring to SimpliVity.
Our partnership isn't an OEM - was never intended to be. It's 100% meet in the channel (ala FlexPod).
I'd hardly call shipping 1000 OmniStack integrated Cisco 240 units in the first year of a partnership 'unsuccessful'.
While we love our partnership with Cisco, OmniStack was/is built to be platform agnostic (see our whitebox, Lenovo partnership, more coming) and that is our GTM strategy.
https://www.simplivity.com/blog/2016/03/coopetition-is-the-new-normal-in-the-hyperconverged-space/
As you well know, Cisco partners with many players, always has. They will continue to partner with EMC as well as SimpliVity.
Cheers to your team, the SpringPath team and to Cisco.
Posted by: Matt Vogt | March 01, 2016 at 01:46 PM