2016 is the year EMC and VMware are going “all in” when it comes to the HCIA marketplace. We took all the hard learnt lessons from the EVO:RAIL experience, and went back to the drawing board in late 2015.
Let’s be clear – EMC is the undisputed leader in the Converged Infrastructure market as a whole. The way we think of the CI market is that while converged/hyper-converged are different system-level designs, with radically different CAPEX/OPEX envelopes. That said – converged and hyper-converged infrastructure are both variations on the same theme: “buy/consume vs. building things that are boring and everyone does similarly, so you can: go faster; invest smarter; simplify how you operate; get better overall availability”.
- We have the leading Converged platform in Vblock and VxBlock Systems – 2nd place is distant, the gap continues to widen. This is not hyperbole, it’s a fact (see here in the customer voice, see here in IDC data).
- There is no peer to VxRack Systems in the Hyper-Converged Rack-Scale platform category. When you are thinking: multiple rack/hundreds of node scales; broad heterogeneity – inclusive of physical, different kernel mode hypervisors, the container/cluster manager ecosystem, the data fabric ecosystem; design/support/sustaining engineering that incorporates the network domain – you’re thinking about a Hyper-Converged Rack-Scale system, not an Hyper-Converged Appliance.
To put it in perspective, there are “only” thousands of hyper-converged Rack-scale customers in the world. On the other hand, there are hundreds of thousands of hyper-converged infrastructure appliance customers.
But – I try to keep Virtual Geek a delusion-free zone :-) It’s not all sunshine and roses. In 2015 it was clear, we needed to go faster in the hyper-converged appliance as we were getting out-run by the startup ecosystem. When you think “Hyper-Converged” most people think start small, start simple, scale easily, leverage your standard – VMware – all of which is so important for SMB, SME, and the Enterprise edge. When you think “hyper-converged”, most people think of Hyper-Converged infrastructure appliances (not the software that powers them, where VMware is doing great with vSphere and VSAN) and not rack-scale systems where heterogeneity, network/SDN are all so critical.
On Feb 16th 2016, we launched VxRail, EMC and VMware’s exclusive HCIA - with a laser focus on the “4 P’s”: Product, Price, Place/Package, Promotion.
- VxRail is unique in a couple of ways:
- …it is the ONLY HCIA that uses vSphere which isn’t developed/engineered and sustained/supported in some “arms length” partnership way. Literally with VxRail is a SINGLE product team, a SINGLE engineering team for the management software stack. That team has one mission: build the best product in that market. Where there is something we need to do better – that team has no place to hide – direct short closed feedback loop :-)
- … it is also the ONLY HCIA who doesn’t have a “first call + handoff” support model. Like all things VCE – we do L1/L2/L3 – there is no “hand off” and we never let go of the case. It’s worth reinforcing that the management/orchestration software that binds hyper-converged systems (appliances and rack-scale systems) is the majority of what defines the behavior of the offer – with the bulk of the rest being defined by the SDS behavior. The hardware itself in HCIA is almost inconsequential - important only insofar as it defines the economic and configuration envelopes (interestingly, in Rack-Scale systems – there is room for hardware differentiation around industry standards – think DSSD, think composable infrastructure – more on this topic soon!). The VxRail management and orchestration software in turn leverages and exposes vSphere and VSAN (whose product management and engineering is in VMware – as of course they are also used outside the OEM VxRail model for those parts).
- … it is the also the ONLY HCIA that has a fully integrated SDS stack that is embedded into the kernel – specifically VSAN because VxRail uses vSphere. No crazy 8vCPU, 16+ GB of RAM for the storage stack (per “storage controller” or even per node in some cases with other HCIA choices!) needed. No need to worry about performance and contention as the hyper-converged nodes are used for what they are planned – running workloads. This is the same reason in VxRack System 1000 with VxRack Flex nodes we use ScaleIO (it is kernel-integrated with multiple hypervisors and physical host OSes).
- VxRail is a different product than what we’ve done in the past (80% different code than during the EVO:RAIL program, and completely different hardware). It is designed and prioritizes all flash configs and IO paths. It is designed for rich data services. It has the highest efficiency and best performance (bar none) of the Hyper-Converged Appliance market.
- VxRail has a totally different price (starting at a $60K ASP vs. almost $200K for our prior efforts!).
- VxRail has different packaging (flexible ability to bring your own licenses, bundled software, completely different distribution and channel).
- VxRail has totally different promotion – with all of EMC and VMware and our partners driving the go-to-market hard.
In Q1, VxRail has been in the market for only 44 days.
It’s been a huge hit.
I can’t say too much – it’s still early days (which means early data points shouldn’t be extrapolated too much). There is a lot for us to continue to prove. Also, we’re in the quiet period before earnings, but I’ll talk about it a bit – while VxRail is off to a GREAT start – reality is that it is still such a small business in the whole that it doesn’t affect EMC’s earnings.
I will say this – in just 44 days, as of March 31st, there are 138 VxRail systems out there, sold in 37 different countries around the world, and the demand/quotes outstanding for Q2 is ENORMOUS. If we are win in even a fraction of those cases, we have the opportunity to move WAY up and to the right stack ranked against other HCIA offers.
Q2 also has a huge update in the software stack (continuing divergence from the EVO:RAIL history), where compression/dedupe/erasure coding get turned on, and the all-flash configurations (which are the sweet spot – and where we are doubling down, with 3.8TB SSDs in Q2, and 15TB SSDs coming later in the year). At that point, I think hybrid-optimized HCIA SDS stacks will increasingly look old-fashioned.
My personal goal is that by 2017, EMC is #1 in the HCIA category by revenue (and IDC will be a good source to validate where we land) – the offers are THAT good.
Today has another big piece of news that will accelerate EMC and VMware as the leaders in hyper-converged infrastructure and hyper-converged software.
Effective immediately, Dell is reselling VxRack Systems and VxRail appliances.
What’s the scoop?
Answer: It’s SIMPLE and it gives customers more CHOICE.
- Structurally, it is a resell agreement between two separate companies, Dell and EMC, similar to other Dell/EMC resell arrangements.
- Strategically it is something good for Dell customers and for Dell. How?
- For Dell customers, it gives them more choice. Dell is the leading channel for other hyper-converged infrastructure appliances. Now, Dell customers have the additional choice of VxRail, the best HCIA for customers who have picked VMware as their standard, and VxRack the best open and flexible rack-scale system on the market. Of course, they have other HC choices in their portfolio – and ultimately the customer chooses!
- It is also good for Dell as a business as it strengthens their position with the broadest hyper-converged portfolio. They can (and will) position the solution in that portfolio that is the most effective for the customer, and wins.
- Strategically it is something is good for EMC customers and EMC. How?
- For EMC customers it gives them more choice. They can choose to acquire through EMC and EMC partners, or through Dell and the Dell partner ecosystems – whatever works best for them.
- For EMC as a business it broadens the reach and accelerates both VxRail and VxRack Systems.
- Strategically, it is something that is good for VMware customers and VMware. How? Again – choice and reach (broadens the routes to market for their hyper-converged software – vSphere and VSAN – which are OEM/embedded in VxRail, but also as standalone software, and as VSAN-ready nodes).
Now, I won’t be surprised if a fair amount of digital ink gets spilled over this news – and I also won’t be surprised if there is some hyperbole (when isn’t there?!)
I suspect some will jump to 2nd and 3rd order conclusions. Advice? DON’T.
This news is pretty simple at it’s core… It is about giving customers more choice via Dell’s entire industry leading hyper-converged portfolio, and broadening the reach of VxRail and VxRack.
Of course I have my bias – and think if you’re a VMware customer looking at hyper-converged – choosing VxRail is a no-brainer. If you are a customer looking at large scale, datacenter, heterogeneous hyper-converged – choosing VxRack is a no-brainer.
Looking forward to getting more customer feedback – and the roadmap (and outlook!) is STRONG!
IMO i think you really need to beef up M&O layer if you want to consider VxRail a no-brainer. Can you provide some perspective on number of units sold? Is that a lot? Don't think you can compare against some others in their first 45 days due to when they entered the market, would love to hear comparisons against Hyperflex or others who have recently entered the market.
Posted by: Josh Coen | April 05, 2016 at 09:43 AM
Just one question. Every hyper converged sold by dell are based on their own servers. What will happen to vxrail and vxrack Which are not based on dell hardware ?? I don't think dell would like to sell other hw than their own.
Posted by: KnThrack | April 05, 2016 at 04:55 PM
@Josh - thanks for the feedback. What specifically do you think we need to add to the M&O layer? There is already rich capability for single click cluster expansion, node refresh, fault and performance management. Would love input.
Re: perspective on unit count. Lemme try. There is no question amongst reasonable people that today Nutanix is the unit count/revenue leader in the HCIA space (though VMware is the leader in software that is deployed in hyper converged forms - both by revenue and customer count). In the updated Nutanix S-1 prospectus, they call out 2638 customers after their roughly 6 years in the market, 494 in the 3 month period ending at Jan 1, 2016. So - for perspective, in the 44 days that we were at it, our volume was somewhere around 55% of the volume of the market leader. I'd say that's a good start :-) Let's assume my math is wrong, dramatically in their favour - and it's roughly at 1/3rd. I'd still consider it a good start.
Ultimately, the milestones are our first 500 customer quarter, then when we cross 1000, 2000, and 5000 customers.
But - I think reasonable people would agree, there's a strong new player in town!
Re: Hyperflex - it's a head scratcher for me. WARNING - I don't claim to be an expert on any competitor.I get VICs and FEX uplinks on traditional converged infrastructure (they are a core part of Vblocks) - but I think that in hyper-converged infrastructure appliances, those add unnecessary cost relative to more industry standard hardware. Also - the core of any HCIA is two things: the SDS stack and the M&O stack. I would poke at the scaling limits, the performance behaviours (particularly with large load on the host). It LOOKS like there's a SDS controller that is pretty big (8vCPU and a lot of RAM) - perhaps someone can confirm/correct. Presumably you need a couple of those controllers. Thats a metric ton of resource utilization.
MOST importantly from a business standpoint - I simply don't get how an OEM (Hyperflex is based on an OEM of Springpath) of the things that actually MAKE the HCIA (the SDS and the M&O) will work. I think anyone who is dead serious about this space needs to invest, develop, and iterate that portion themselves. It's certainly our approach - we think we must own the responsibility of innovation and iteration in the SDS and M&O domain.
I think if Cisco was really serious in HCIA, they should go ahead and buy Springpath. I think spring path also has a very small number of customers in the installed base, and is a relatively new SDS. I'm sure their engineers are very bright, but it takes time for any storage stack to harden - and I hate to say it, but you only get that with volume of customers.
I don't know who else you were talking about, but if you ask, I'll share my thoughts (biased as I'm sure they are, but no baloney, will shoot straight).
Clearly the weak point of VxRail is simple. It's designed totally for vSphere. The following statements are true clear and sustained FOR vSphere focused customers: There will be no HCIA more integrated with vSphere. There will be no HCIA with a more clear support model - end to end inclusive of the hypervisor with NO HANDOFF, PERIOD. There will be no HCIA with a better performing SDS stack with vSphere. There will be no HCIA with a better VM-level policy model.
There will be customers who are all about Hyper-V. VxRail is not for them.
There are people who will be all about KVM. VxRail is not for them.
But - if a customer has standardized on vSphere - VxRail is a no-brainer.
Now, I'm SURE that competitors will put the fear of god knows what about "lock in" - but that's total hoo-ha.
If you pick Nutanix, you're "locked in" to NDFS, Prism, and you know they will push toward Acropolis.
Frankly, with any appliance, you're locked into the appliance for however long you have it. That's no better/worse that VxRail. But these days migrating on and off hypervisors and HCIAs is easy enough - that "lock in" at that layer is FUD (in my opinion). Compared to the lockin of a public cloud, or SaaS (!!), there's no lock-in at the infrastructure domain anymore.
Thanks for the question!
Posted by: Chad Sakac | April 08, 2016 at 02:49 PM
@KnThrack - thanks for the question!
There's no question that we're going to leverage the snot out of Dell's supply chain IF/WHEN the Dell/EMC merger goes through. We would be crazy not to!
1) it will lower the system cost.
2) it will broaden the range of VxRail platforms (1U1N, 2U1N, broad GPU support, etc)
3) it will be a more flexible supply chain (more build to order)
4) it will be a faster supply chain - it's no coincidence that Dell XC was the first Nutanix platform to support the latest Broadwell CPUs.
Dell is wicked awesome at 1, 2, 3, and 4. Of course, this isn't a offer from the combined company. It's two arms-length partners (Dell and EMC), and Dell is reselling VxRail and VxRack as they stand.
So, I think it's very fair to conclude that over time, should the Dell/EMC merger go through, we'll accelerate the offers with Dell based hardware - but VxRail and VxRack are compelling as they stand, and for customers that want them, and want them from Dell, they are now able to go that way!
Thanks for the question.
Posted by: Chad Sakac | April 08, 2016 at 03:31 PM
VxRail looks great, will be giving it a look in a couple of months. I just have a couple questions. I thought I read somewhere else where you will be offering CIFS/NFS/ISCSI on top of the VSAN storage. If so any approximation?
Second questions is limits of VSAN. It looks like a great product however one thing that holds me back on it is snapshots. The array vendors and HCIA can take hundreds/thousands of snapshots based upon schedules and keep them for days, weeks, months. To my knowledge VSAN can take a maximum of 32 per VM and I dont know if these can be scheduled or not either. I feel that this is a weakness compared to others and I would personally really like to see VSAN offer thousands of snapshots with scheduling. Can you comment on any down the road enhancements or if there is even any consideration to this?
Posted by: JR | April 08, 2016 at 08:23 PM
** Disclaimer: I work for VMware in the Storage & Availability BU **
I cannot publicly comment on roadmap items and commit to timelines. I can however say that we are working on improving the current snapshot technology. I cannot imagine however that even in that case we will be talking about 1000s of snapshots of a given VM. Personally I have not encountered customers either with a use case for snapshots at that scale. I would be interested in knowing what your use case is @JR, feel free to drop me an email on [email protected] to discuss this. 1000s of snapshots becomes unmanageable really fast, even dozens per VM is very challenging, so would be interested in knowing what you would do with this capability.
Thanks.
Posted by: Duncan | April 09, 2016 at 04:09 AM
Thanks for the follow up Duncan and Chad. I have emailed you at your provided address. This is just for others, not necessarily 1000 or more snapshots per VM but for all VMs in totality, certainly more than 32 per VM though.
Posted by: JR | April 10, 2016 at 01:37 AM
VxRail looks great for VMware only shops and I am sure a significant % of newVxRail implementations will be integrated with Brocade VDX switchs over time. I have a question re. VxRack, are there any potential plans to integrate Brocade Data Center switchs & SDN solutions in the future with VxRack, as an alternative option to Cisco? IMO this would have significant appeal to many Enterprise/Data Center/IAAS/PAAS shops!
Posted by: Russell | April 18, 2016 at 06:36 AM