Building a Predictable and Cost-Effective Virtual Desktop Infrastructure (VDI)

Virtual Desktop Infrastructure (VDI) is a desktop-oriented service that hosts user desktop environments on servers and/or blade PCs. I recall having conversations with clients 4-5 years ago around VDI, when many predicted that it’d be the next big thing in IT efficiency. All of the efficiencies of server virtualization (shared compute & storage resources, ability to provision/de-provision on demand, simplified patch management, monitoring and support, etc.) could now be easily applied to the desktop environment. It was the hot topic of 2010/2011, everyone was talking about it but very few truly jumped on-board mainly because of cost considerations. A full-scale VDI deployment still required expensive assets: OS licenses, VMware licenses, servers, storage, etc. While we could illustrate anticipated soft cost savings, i.e. operating expenses such as ongoing management and administration, it was difficult to justify in lieu of hard dollar savings. Ability to scale and maintain performance was also a challenge, given the VDI environment comprised of physically discrete infrastructure of compute, storage, and network.

Fast forward to today, clients are not only talking about desktop virtualization, they are rapidly adopting it across the enterprise. We believe this is largely due to the proliferation of converged and hyper-converged infrastructure, and players such as Nutanix have completely changed the game. With compute, storage, and network now fully integrated (in a distributed manner) along with ‘single pane of glass’ UI that provides an integrated analytics dashboard for the data center, Nutanix has enabled the customer to truly have a software defined architecture, which is ideal for a VDI deployment. It is also a turnkey solution that comes ready to support desktop brokers, such as VMware Horizon View and Citrix XenDesktop, as well as supporting toolsets such as LiquidWare and Unidesk. Finally, this virtual computing platform would support every type of VDI user—from task and knowledge workers to power and data scientists. This unique platform and architecture provide customers the simplest path to enterprise deployment with the agility of public cloud providers.

The Nutanix infrastructure is a scale-out cluster of high-performance nodes (or servers), each running a standard hypervisor and containing processors, memory, and local storage. Each node runs virtual machines just like a standard virtual machine host (please refer to the diagram below).


This out-of-the-box infrastructure solution eliminates high cost and the extensive risks posed by conventional virtual desktop solutions. Customers can minimize their upfront cost by starting small (as small as 3 nodes in a single 2U appliance) and adding nodes as they grow (scaling from 50 to tens of thousands of desktops), maintaining absolutely linear and predictable performance. Clients also reduce risk by leveraging pre-validated reference architecture. Nutanix offers a ‘Virtual Desktop Bundle’ offering that allows clients to purchase a fully baked VDI solution based on the number of users.

In addition to hyper-converged infrastructure, we’ve noted other trends as Desktop Virtualization continues to gain momentum in 2015:

  • Blurred lines between Virtual Desktop Infrastructure (VDI) and Bring Your Own Device (BYOD).
    • As usage on mobile devices continues to proliferate, attention is turning towards mobile device management (MDM), security and support.
    • MDMs need to step up to ensure a consistent management experience across multiple types of devices (ie. PC, Google Android, Apple iOS).
  • Operating System (OS) irrelevance.
    • Clients are reluctantly being forced into Windows 8 and, with thin client computing becoming more feasible, the OS becomes irrelevant to a certain degree and users have choices (ie. Linux, Google Chrome OS, etc).
  • Solid-state storage for VDI deployments.
    • Traditional vendors and new vendors (ie. Nutanix, Nimble, and Pure) are gaining market traction – delivering unlimited IOPS per desktop solves a major pain area for VDI deployments.
  • Open source VDI.
    • Early stages project underway to deliver VDI desktops from an OpenStack Cloud.
    • Hypervisor license costs may not be a huge cost of a VDI deployment, but every opportunity to save money helps!

We’re delighted to see technology transformations, such as hyper-convergence and web-scale technologies, that allow VDI to finally take off and deliver value to the enterprise! It will be interesting to see the continued evolution as momentum gains around open standards, which will deliver greater flexibility and choice to both the business and the end user.

Arrow-InforaphicHyper-converged is not just a reference architecture re-wrapping the servers, storage, networking and hypervisors you use today. Hyper-converged is truly a converged platform, meaning servers and storage packaged together with a network connection. It becomes web-scale when you add software defined storage elements into the platform. Think about Google, Facebook and Amazon – they could not run their services effectively and efficiently if they were bound by traditional IT infrastructure.

Today I’ll talk about how to identify if hyper-convergence is a fit for your organization.

If you are struggling with performance, cost and complexity as you deliver IT services, you are not alone. These three reasons are key underpinnings for looking at hyper-converged or web-scale infrastructure that allow you to recalibrate your approach to IT service delivery. Why? Let’s look at all three.

 Performance: The acceleration of applications by moving from disk to flash has helped clients unlock some additional value in the infrastructure. However, any IT architect will tell you flash is not always the answer – in fact, you may be trading one bottleneck (I/O and latency) for another (overloading your network or taxing the CPU’s cycles). Hyper-converged/web-scale solutions provide the best of both worlds. The solutions are pre-packaged to leverage flash and disk backed with a software defined solution. You will get predictable performance for your applications and enjoy the benefits of compression, dedupe, tiering between flash and disk and software defined RAID. Most clients have to pay extra for these features and as a result, many storage solutions in the data center today have one or two of these features, but not all of them. More important is how these features are packaged to work in tandem, providing exponential performance gains in a scenario of 1+1=3.

 Cost: If you are not trying to find a way to reduce CAPEX or OPEX in your IT budget, you are unique and should enjoy it while it lasts. I meet with 200 clients a year and I can tell you cost is in the top two for IT decision making. The question being asked throughout the organization from CEO down is how do you reduce costs while delivering IT services competitive with SaaS and cloud based IaaS providers? There is no wrong answer here. In the context of hyper-converged/web-scale infrastructure, you pay as you go. Start with what you need and scale up. Pick what you need to keep the infrastructure balanced. When you compare cost models for traditional infrastructure to these new models you can save at least 20%, simply because you are combining servers and storage into one package. Operationally, you no longer require a server and storage admin, tasks can be handles by one resource.

 Complexity: I love this one, because we all live in this world. What if you could get a better handle on the complexity of delivering IT services?  Instead of having racks and racks to support an environment, what if you could condense that into a single rack, or even a half a rack, but deliver predictable performance that is reliable while (as I mentioned above) saving 20%? There is not a CIO in the world that would not sign-on to have that in place. Moving storage and servers together is not new. We had application servers with direct attached storage and before that mainframes. What IS new is the innovation in software development that enables everything to work in harmony as you scale.

Let’s talk about specific use cases:

  1. Virtualized infrastructure refresh
  2. Private cloud PODs
  3. Remote Office
  4. AWS/Softlayer today but looking to buy infrastructure soon
  5. VDI

If you have questions or would like help in determining if hyper converged is a fit for your organization, give us a call.


It seems that this new term, hyperconverged has come screaming on the scene in the last 6-9 months. Why? Is converged infrastructure as a solution architecture just not good enough anymore? Checking in with analysts and clients, the converged systems or converged reference architectures are being deployed at a rapid rate, in fact Gartner said the space was going to be a 32% CAGR from 2012-2016 versus traditional server purchases at 5% CAGR. Analyst data aside, the converged market is booming at a healthy single digit billion dollar run rate and primarily driven by VCE, Flexpod and IBM. It sure seems the converged system space is very healthy, growing and not going to be in decline for the foreseeable future as enterprises get their arms around optimizing their IT environments for the journey to the cloud. So what exactly is hyperconverged and why is the industry buzzing about it?


Hyperconverged, simply stated is the squeezing of compute, network and storage into a single package with a single pane of glass to manage. AHA! That is converged systems story, I know you’ve heard the story already. There is a difference here. The solution truly is a single package. When you compare this to VCE and Flexpod, those solutions claim to be “single packages”, but they are nothing more than a reference architecture representing a collection of technologies that are pre-integrated. These solutions are not small in scale either, often times we are talking about taking up as little as ½ a rack that is 42U tall. The hyperconverged solutions are 2U of rack space, not ½ a rack. That is one difference. The other big difference when compared to other convergence offerings in the market, is that these hyperconverged solutions can be clustered together and expanded. This allows you to achieve performance, resiliency and storage growth in a plug and play fashion.


So a quick recap. Hyperconverged solutions are smaller packages, built for scale. All of the same features and benefits of converged infrastructure/reference architectures still apply. I believe this fact, coupled with the entry price for a solution starting at $22,000 is driving this new segment in the industry. Hyperconverged is like democracy, everyone can now have a vote and buy in. Converged Infrastructure has been a very large enterprises or rapidly growing companies that are becoming large enterprises. Hyperconvergence is allowing small mid-market companies to jump on board. It is this bottom up approach that makes the market exciting. Hyperconverged solution will continue to push into large enterprise, in fact some large enterprises have been making investments already and adoption will continue at an accelerated pace.


Hyperconverged solutions may appear to be a buzzword, but they are in fact reality. Nutanix has made the market for all other players. Simplivity has now jumped in the game as well. Vmware, EMC and Cisco are all working on hyperconverged solutions. That is the signal that we are now in a maturing mainstream marketplace. Next month we’ll take a deep dive look at Nutanix. We are a big believer in their technology, ability to execute in the market and impressed with their client support. All key ingredients that must be in place for CMI to represent a brand in the marketplace.