VxRack SDDC–VxRail’s “Big Brother”

[updated 12/13/2016 @ 8:30amET – see below]

Yesterday, VxRack SDDC got a big update – like VxRail and VxRack FLEX before it, reborn on Dell EMC PowerEdge.

“Big Brother” isn’t a reference to scale (VxRail can scale as big as you want, in multiple clusters) – rather more a case of a more complete SDDC stack – inclusive of networking.   The domain of a single VxRack SDDC system can be 190+ nodes, spanning more than 8 cabinets, with many internal workload domains.

Our top secret plan isn’t so top secret – so I’ll lay it out here:.

1. I believe that HCI (along with CI – a bigger market, but growing more slowly) is the future of the on-premises infrastructure layer.    The monstrous HCI growth rate suggests this is right.
2. I believe that HCI is all about software – but that the dominant consumption form is in system form (software/hardware together).  Why?  Simple – people want simple.   You cannot have turnkey outcomes without a system approach – PERIOD.   It’s no coincidence that hyper-scale players use “software defined” approaches with their own open approach to hardware… but yet maniacally design and operationalize the two layers TOGETHER as one (and have vast “hardware as a service” teams).  No hyper-scale player uses HCI Appliances or Rack Scale systems – but in effect they are building their own.    For anyone who isn’t an hyper-scale provider and doesn’t have a vast “hardware as service” team, HCI systems is for them.
3. I believe there’s a critical idea that flows from #2 – which is that to win in HCI – while it’s all about software, ultimately to win, you need have a massive x86 server business, global supply chain, global support model, and even go so far as designing the hardware building blocks to optimize for SDS and HCI.   We do, and we will be doubling down on that path.
4. I believe that HCI systems come in two main system-level architectures – HCI Appliances (which are all about starting small, as simple as possible) and Rack-Scale systems.   Appliances leave ToR physical networking and SDN out of scope.   Rack-Scale systems have ToR physical networking and SDN at their core.    BTW – the number of HCI Appliance customers are 10-100x more numerous than HCI Rack Scale… but the market for each (in dollars) is about the same.
5. I believe that there are two approaches for HCI Appliances and Rack-Scale Systems:  a) vertical “stackification” (where the whole stack lines up with VMware, or with Microsoft, or another stack – all the way up to the IaaS/PaaS layers); and “generalized HCI” which provide automated, pooled compute/network/storage – but you bring your own compute abstraction stacks.   The models with complete vertical stack integration will ALWAYS be simpler/easier to operate and lifecycle – because they take the standardization several steps further (that fundamental trade-off of “flexibility” and “outcome” manifesting again).

I firmly also believe in a business principle: to build kick-a$$ offers, you have to have one team, clear mission/charter, and accountable to the field/partners/customers.   If you do that, and iterate fast, you get good, FAST.    This is true in big companies just like it is for startups.

The formula for the spectacular reboot of VxRail (which is now a 9-month old toddler taking over the HCIA market) is the same one we’re applying to VxRack SDDC.

What’s that formula?  SIMPLE. 

• We stopped messing around, built one team – made up of Dell EMC and VMware folks working as one, we have started to apply that approach to VxRack SDDC.    In fact, it’s a single team – that lives and breathes the VMware stack. 
• That team is singular, and maniacal.   Their mission is to make the best HCI Appliances and Rack-Scale systems for customers who are all about VMware and the VMware SDDC as their standard stack.  
• They stay in lock-step with VMware SDDC and the Cloud Foundation stack – and that includes things like VMware Integrated OpenStack (aka VIO) and vSphere Integrated Containers (aka “VIC”) – which is now GA (see more here).
• We invest (massively – this is code named “Project Merrimack”, and is measured in many, many millions of dollars), set massive goals, apply the full force of the whole company, measure success/failure and then iterate like crazy.

I’m not a genius – this is a pretty basic formula for success :-)

We have other HCI Appliances and Rack-Scale systems for customers who are not about the VMware SDDC but rather want to take a “workload abstraction neutral” approach – namely Dell EMC XC built in partnership with Nutanix and VxRack FLEX.   It’s “round one” in the HCI era – and it’s also early enough that HCI is not a “zero sum game”.  

But let there be no doubt:  if you’re standardized on vSphere – VxRail and VxRack SDDC are your best bet – PERIOD.

[update – 12/13/2016: 8:30am ET] – some people have asked me: does this mean by inference that you are saying the storage domain for vSphere is vSAN – always?   NO – and that isn’t the position of VMware, or Dell EMC for that matter.    What it does mean is that VxRail and VxRack SDDC are the best bet to start the dialog – and for many customers (most) are the best HCI answer from Dell EMC for most vSphere customers.   Also, we are indeed finding that the majority of vSphere workloads (of many types) by COUNT (a percentage) are happy as a clam on vSAN – but it’s not that simple, not that absolute.   If you want to virtualize everything (and at all customers) – you need a broader aperture.  

Workloads vary.  Customers vary.   Anyone thinking that one stack is enough for the market is wrong.  Don’t trust people who don’t ask you some questions before giving you an answer.  

One thing that people don’t generally know is that VxRack SDDC can incorporate external storage (DELL EMC VMAX, XtremIO, Unity, SC series, and also ScaleIO) for the workloads that are not a fit for vSAN for one of several reasons.    We’re taking this idea much further, much faster throughout 2017 – if we didn’t do this, it would limit the usefulness of VMware Cloud Foundation and VxRack SDDC to a subset of the use cases.   I’ll do a longer blog post series on this in the new year.

And… wait for it – HCI is all just a side-show.  The main event is making turnkey IaaS/PaaS/Analytics platforms possible.  

I was listening to Episode #233 of Speaking in Tech where they were talking about the VMware SDDC on AWS announcement and paraphrasing: “the existence of SDDConAWS is an indictment of on-premises IaaS”.  They are right.  

On-premises clouds need to become radically simpler, easier to deploy/consume.   They need to have radical simplification of the lifecycle management process.   They need to scale as you need, and they need to have similar economic models as the public cloud.  PERIOD.  And I’m saying that as the leader of the team at Dell Technologies that has deployed more successful Enterprise Hybrid Clouds on VxBlocks on anything else.

It’s a great start – but we must do better.

This is our “Apollo moon mission” for 2017.

We will use VxRack SDDC (and VxRail) as the basis to radically simplify the Enterprise Hybrid Cloud.   It will mirror the utility economic model of SDDConAWS.  

Oh – and clearly there is another emerging “stack” that customers want the same approach.   We have a parallel team (clearly isolated from the VxRail/VxRack SDDC team) with a similar singular purpose and mission: to build VxRack Azure Stack in a unique and differentiated way, and kick-butt there.   How will we do it?  Simple.   One team, completely aligned with the Azure team at Microsoft.   Maniacally focused.  Invested.  Measured.  Consumption models that map to the off-premises pairing.   Then we will iterate… and fast.

Oh – and we will make Pivotal Cloud Foundry run on each of these – on and off-premises.

We probably have two more of these “very tightly integrated vertical stack on an engineered system” left in us and then I think we’re complete.

Being transparent as always: it is my intent to use HCI Appliances and HCI Rack-Scale systems as the basic building block to make the dream of simple and easy hybrid cloud models, and in doing so, I think we at Dell Technologies have a chance to do something completely, totally awesome.

Comments

[Craig Bruce]

Hi Chad,

I always enjoy reading your articles... whilst wordy they are always a good use of time IMO.

You mentioned "On-premises clouds need to become radically simpler, easier to deploy/consume. They need to have radical simplification of the lifecycle management process. They need to scale as you need, and they need to have similar economic models as the public cloud"

EHC has been taking us in this general direction for some time, but neither the EHC or SDDC stack are what I would call "simple" can you elaborate a little on how we are going to accelerate the journey to private cloud? For example, when public cloud IaaS capability is upgraded (new software/hardware/catalogue capabilities)the consumer isn't even aware that it's happening ... "it just happens"

The reason i ask is because typically deployment is the easy bit. Feeding, watering and growing + exposing a consumption model (a usable/fit-for-purpose service catalogue that can easily be consumed) is the most challenging element (essentially life-cycle management of the private cloud IaaS) without touching on organisational challenges.

Dealing with brown field deployments (micro segmentation etc) and legacy is another interesting area... (no magic bullet..). Are we doing anything here (apart from NSX) to make this easier?

Also hardware consumption models seem to be very challenging for transitional product vendors.... they typically fall into a "lease type" arrangement.

Thanks.

Craig.

[vikrant]

Is it possible can we use VxRack SDDC for disaster recovery solution ? Please reply.

[taliz]

Seems Dell EMC are still not ready to actually deliver this.
According to VMware they're not done with SDDC Manager yet.
Really unfortunate.

[Chad Sakac]

@Vikrant, yes, you can use a VxRack SDDC in a replicated DR scenario - using either native vSphere Replication, or if you need something a little more enterprise-y, Recoverpoint for VMs.

@Taliz, that's not quite accurate. VxRack SDDC 1.0 is released, and is being sold. What is happening through the whole of 2017 is an aggressive roadmap to do a ton of things: bring more hardware variation into the platform (mirroring VxRail), integrating and extending SDDC manager above and beyond VMware Cloud Foundation (which will always be the core). This will add things like ESRS ("remote support connectivity"), things like the data protection and DR capability we've done on VxRail, and lots of the lower-level hardware telemetry and fault we've learnt is necessary. There's actually even more to it - our aim will be to also be able to have external storage better integrated as an option where vSAN isn't enough for whatever reason (though as noted in my SDS blog, this is a smaller subset, but VERY important workloads), as well as being able to lifecycle ScaleIO in workload domains.

Drop me an email - would love to know where you heard that, as it's wrong, and I'd like to correct them. Thanks for commenting!