Your Resource for All Things Apps, Ops, and Infrastructure

Dell EMC World 2017: Dell HCI Overview

As the Converged (CI) and Hyperconverged (HCI) infrastructure practice lead at AHEAD I spend a significant portion of my time evaluating and designing solutions that include Dell CPSD components (CPSD stands for Converged Platforms and Solutions Division – formerly VCE). My last post focused on VxBlocks and Vblocks, providing an update on the products themselves, as well as providing my opinion on how I expect these CI products to fair during the HCI “frenzy” that we are currently in.

This post is going to focus on the recent HCI announcements from the inaugural Dell EMC World hosted in Las Vegas last week.  I will also touch on some of Dell’s competitors. Not-surprisingly, HCI products continue to be messaged as an on-premises solution that offers public-cloud value.

Let’s dig into recent technical and non-technical developments and decide for ourselves.

Technical / Product Updates

I don’t like regurgitating technical specs that are available in manufacturer presentations but it is useful to understand the basics of the recent announcements. If you already know this, skip further down for my opinions.

VxRail 4.5

  • HW refresh onto the much anticipated 14G servers (Skylake-SP)
  • VSAN 6.6 (skipping 6.5) and vSphere 6.5 support, bringing with them:
    • Data at rest encryption
    • Enhanced Stretch Clustering (the prior implementation was not good)
    • No more Multicast (no more wildly unnecessary and ultimately un-impactful conversations with concerned network admins)
    • Improved alerting (crossing my fingers here, VSAN is known for being excessively noisy when it comes to alerts)
    • Access VSAN management directly through an ESX host (great for vCenter outage situations)
    • Improved HW monitoring and proactive protection, leveraging Intel SMART
  • Broader configuration options (not quite CTO, but close), specifically with expanded options around single socket systems, and more network port options
  • Replacement of VMware Data Protection with Avamar Virtual Edition (good since VMware canceled VMware Data Protection)
  • A focus to reduce interaction with VxRail Manager, in favor for moving more functions to vCenter
  • Availability – September 2017

XC (Nutanix)

  • XC430 Xpress, a 3 node configuration that is aggressively priced (similar to the NX series Xpress program)
  • Bundling of Dell data protection offerings (Avamar virtual Edition, Data Domain Virtual Edition)
  • Availability – June 2017

VxRack (ScaleIO & SDDC)

  • Support for E7 nodes for the FLEX variant opens up more intense workloads to this platform along with the ability for software publishers like Oracle, SAP, Epic, and others to consider approving this platform for some of their products
  • Broader configuration options for SDDC variant (more CPU, RAM combinations).

Technical / Product Updates – Opinion

Vxrail Focuses on Incremental Updates – VxRail 4.5 updates predominantly focused on making the entire experience better instead of focusing on new features, in other words, an incremental improvement. The 14G hardware platform will speed things up and ensure that the VxRail platform leverages the newest hardware. Access to VSAN 6.6 (albeit 5 months after it’s release) will further provide performance, monitoring and manageability improvements. It will also unlock a few strong features (encryption, improved stretch clustering) although these are arguably approaching “checkbox” status. The removal of multicast is fantastic because it removes a big concern for network managers (what was an unfounded concern given the way in which VSAN used multicast)

Incremental is GOOD – Don’t take my comment as a negative: making VxRail a better product at the risk of not introducing revolutionary features is OK. Dell EMC’s portfolio already has many overlapping products so the approach of making VxRail better, and then combining VxRail with other products (data protection, scale-out NAS, etc) is a very strong strategy. VxRail’s primary weakness is that it isn’t 100% baked (and to be fair, very few products are).  If Dell can make a flawless VxRail with the features that it already has, it would be a tremendous force in the market.

What I would like to see on VxRail (my wish list)

  • That the noisy alerts go away (this is more of a VSAN thing than a VxRail thing)
  • That the confusion over Log Insight licensing is resolved (You get it for free, but only if you use the internal vCenter license, which most customers don’t use)
  • That Enhanced Stretch cluster is a viable active-active solution
  • That VxRail Manager and all the vestiges of Marvin are completely hidden from the end user and that the solution is end to end managed through vCenter, without any concerns when renaming clusters or moving clusters to different vCenters.

Nutanix XC Updates Opinion – Meh. That is my feeling on the updates from Dell EMC World on the XC platform. And perhaps that was their intent, Although Dell enjoys a large pipeline on the XC line they would rightfully much rather sell VxRail which has 100% Dell content. I do have some thoughts on Nutanix’s consumption model for their NX line which I talk about later in this post.

VxRack Updates Opinion – Good. They addressed two of my top three concerns and are working on the third. E7 support for VxRack FLEX makes it a viable platform for a broader set of workloads, specifically the tough ones like memory hungry (2TB+) databases. The increased node options for SDDC is critical, up to this point it was nearly impossible to fit a workload to a VxRack SDDC because they only had a couple of CPU/RAM options. I couldn’t attend EMC World but I did see that some sessions were published around Symphony which is a project intended to create very powerful automation and management toolset for VxRack which the product needs.

Consumption Updates

The Register published an opinion lambasting Michael Dell for not having a cloud strategy. I’m not sure whether I agree or not on that, we have seen several companies attempt to invest in a cloud offering and fail, but on the other side, we continue to see public cloud and SaaS offerings continue to expand, at the expense of on-premises manufacturers like Dell. Well, if you don’t want to be them, copy them… Behold Cloud Flex.

Dell Cloud Flex – A rent program for VxRail, where you get more gear on-premises than you need, but are only charged for what you use. You have a 1-year obligation after which you can return the gear with no penalty. The more you consume and/or the longer you consume the cheaper it gets. This offering is currently available only for VxRail but it is expected to be expanded to more products (where it makes sense). The idea is to always have excess gear in the data center, so you can scale up immediately when needed.

Dell Cloud Flex – Opinion – Cloud Flex is an attempt to mimic the on-demand, 100% opex, minimal obligation, reduce cost as you eat more, consumption model that is table stakes for public cloud providers but non-existent for on-premises strategies. Many IT providers have previously attempted to get this model through complicated leasing vehicles, most of which was smoke and mirrors.

Fundamentally the cost of the hardware and software exists irrespective of how you want to pay for it. If you decide to consume an IT asset for 1Y but that asset has 5Y of life, either the consumer or the manufacturer will pay too much or not charge enough. It’s too early to tell but it looks like Dell is willing to take on the risk of not charging enough, with the idea that customers will consume enough of it, for a long enough period, to make this a profitable venture. The question is whether Dell has a sound strategy on what to do with all the gear that is returned back to Dell prior to its end of useful life. Another question is how they are operationally going to track and charge customers for nodes that are online, vs nodes that are off.  Of interesting note to partners like AHEAD, we are also curious to understand the economic model for AHEAD to assist our clients in acquiring and deploying technology in this fashion either standalone or as a larger part of an enterprise cloud strategy.

What is the rest of the market doing?

  • Nutanix – Nutanix Beat Dell to the punch for a consumption (rent) model with their announcement on May 3. The program, named “Nutanix Go” gives customers the ability to engage in short-term (6-12 months) or longer term (2-5 year) agreements. Note that this program is only available for the Nutanix NX line (their flagship) and is not available yet for Nutanix OEMs like Dell or Lenovo.
  • Cisco – I haven’t seen or heard of a consumption model offering but they did release HX 2.1 in April which introduced a competitive offering for ROBO (3 nodes, 1GB networking, no FIs) along with SED support

Wrapping it all up

After digesting the move by on-premises manufacturers to offer rent programs for HCI nodes (which are intrinsically better suited to this program vs, let’s say a giant storage array) I’m reminded by a recent article that I read about decision fatigue. The article commented on the daily routine of several known leaders in politics and business and highlighted that they embraced monotony in their life. Surprising, right? The idea is that making decisions, even simple decisions like what you eat for breakfast, drain a little bit of mental energy, and create a little bit of mental fatigue. These leaders take it to an extreme by wearing the same types/colors of clothing (blue suits, every day), they follow the exact same morning routine (wake up time, exercise, breakfast) to avoid making decisions so that they are as alert and refreshed as possible and able to tackle the real challenges of the day.

I feel like the HCI market, especially now that the rental programs are released, is doing the same thing. The rental programs, if they prove to be priced correctly, makes infrastructure decisions simpler and more importantly, it reduces the number of decisions that need to be made. Once you establish a cost per CPU/GB/TB, which these programs directly do, the decision is automated. Just enter your required CPU/RAM/Storage into the calculator, and there is your cost. Place an order. Plug it in, click a button that automates the cluster expansion or net-new cluster creation (all HCI goodness) and your infrastructure has been expanded to meet business demands.

All the major players (Dell, Cisco, Nutanix) already have taken it to the next step, by automating upstream tasks (App/VM/Container deployments) with solutions like vRealize, UCS Director or NuCalm (Calm.IO).  

With these consumption models in place combined with HCI automation the typical infrastructure manager won’t even be involved (gulp)! I guess it’s time for us to focus on more important IT challenges.  

Subscribe to the AHEAD i/o Newsletter