The Stack Market: Vblock, FlexPod, VSPEX (and me)

I’ve written several times on this blog about converged infrastructure and a few recent events have compelled me to revisit the topic. These events include the launch of EMC’s VSPEX solution along with my decision to accept a new position with a NetApp/FlexPod partner.

I’ve spoken very favorably in the past about the Vblock and even interviewed EMC’s Chad Sakac on this blog about the Vblock, the value proposition, and how it compares to other “stacks”.  Have my opinions changed? And what does the introduction of VSPEX mean for the market? Let’s start by looking at stacks using a car analogy.


Imagine that you are a sports car enthusiast and you have three different choices for obtaining a vehicle. One option is to pick out the best individual components and build your shiny new custom sports car from the ground up. You get exactly what you want from whomever you want, but you have to incur the engineering burden of figuring out how best to get all those parts to work together —  and if you change a part or two in the future you’ll have to re-engineer it then as well. This build-your-own approach is the traditional way of doing things, but it requires good amounts of both engineering expertise and time.

Don’t want an off-the-shelf Corvette? Customize it!

Now there’s one company out there who has selected best-of-breed components from multiple vendors — the best chassis, tires, engine, transmission, and more — all engineered to work together to form one fantastic sports car. This car can then be purchased and supported as a single product. Simply select the size and model you want and you’ve got one great sports car ready to burn some rubber.

But what if you had a favorite gearbox or set of tires that you wanted to use in order to change your ratios or to achieve some other specific goal? You’re kinda stuck. Now depending on your goals that might not be such a bad thing, but some might want a more middle-of-the road solution that offers more flexibility.

Some may prefer to step into the fashion stylings of this custom car

Fortunately, there is a second company that has put together a reference architecture. In this model, the company has selected a finite list of best-of-breed solutions and has certified these solutions to work together to achieve the desired results. Like the first solution, the engineering burden is removed from you the customer, and you’re getting a solution that was designed to work together. In addition you have a bit more freedom to pick best-of-breed components, but you are forgoing some of the benefits of the single product model such as procurement and support. In short, you gain more flexibility while taking a step back from the “pure product” concept.


The Vblock is of course the “product” solution in the example above. It’s an excellent solution comprised of EMC Storage, VMware vSphere, Cisco server (UCS) and networking hardware, and Unified Infrastructure Manager for orchestration across these components. The FlexPod uses Cisco UCS and Nexus (as does the Vblock) but offers a choice of hypervisor and leverages NetApp storage solutions. FlexPod customers can also select from a variety of orchestration solutions ranging from Cloupia, CA Automation Suite, Cisco IA, and many more.

This stack is comprised of three components designed to deliver a tasty breakfast

For a variety of reasons the FlexPod has been catching on as of late enjoying more than 400% growth in customer adoption in the past year. While the Vblock has experienced a measure of success, it has become clear that for reasons ranging from the technical to the political, there will be a market need for more “flexible” solutions –as opposed to the single product model — for some time. EMC reacted to this and the success of the FlexPod by launching their own reference architecture called VSPEX to fill this gap in their portfolio.


There are certainly some for which the product approach is a good fit, and others will opt for the relative flexibility of the reference architecture model. Now everyone with an agenda will try to give their own spin on why reference architecture X or storage solution Y is better so here’s my opinion/spin:  they are all great solutions — you need to find the right fit for your business and technical goals.

We can split hairs and compare EMC Storage to NetApp storage and I’m sure each will find what they believe to be relative advantages. As for VSPEX, yes it does seem to cover a broader array of hardware than the FlexPod, but I happen to be very comfortable with Cisco UCS blades just like the Vblock solution is. And with so many strong management solutions available (Cloupia, CA, BMC, Cisco IA, etc.) it becomes easier to tailor the solution to your needs either as a solutions provider or a IT organization.

A sample FlexPod for VMware configuration

In several ways I believe that EMC’s decision to introduce the VSPEX is a validation of the success of the FlexPod solution which has been extremely well received in the channel.  Some will of course espouse that the VSPEX is somehow superior because it includes more vendors and components, but I think that at a certain point the value of a solution can be diluted.  The components of the FlexPod are certainly best-of -breed including Cisco UCS, VMware vSphere (or Hyper-V), NetApp storage and then customers can select the orchestration and management elements that best align with their needs and goals.  In fact, I really don’t need to state that I think that the FlexPod is a great solution — the clear market success combined with EMC’s decision to attempt to emulate this with their own reference architecture (VSPEX) speaks for itself I think.

When we take our propeller-hats off and take a step back we see that the real value here to the customer isn’t component X, Y or Z, but rather the sum of all of parts in being able to provide private cloud and IaaS solutions which provide captial, operational and strategic benefits to the organization (see “What is Cloud Computing?“). I couldn’t be more excited to have the opportunity to work with NetApp and FlexPod solutions that have been proven to be remarkably successful and have delivered phenominal value to their customers.


Your business has a new initiative which requires a web farm of 10 servers, plus an additional 12 servers supporting databases and middleware.  How long would it take to design, purchase, engineer and then provision the resources for this new initiative?  Using the traditional “build-your-own” approach this could take months, significantly impacting the success and trajectory of your business initiatives.  A well designed converged infrastructure such as a FlexPod solution can provide the technical foundation for vastly reducing your “mean time to provision” and improving your business and profitability as a result.

How is it possible that companies like Instagram and DropBox had only 13 and 89 employees respectively at last count?  Who maintains the datacenter, the cooling, the cabling, the halon system, and everything else that goes along with a datacenter?  The answer is that they leverage the cloud.  In this same manner, mid-to-large companies can leverage IaaS using solutions like the FlexPod in a variety of cloud scenarios to achieve the same type of success.  The FlexPod is an outstanding platform towards this end and I’m very excited about working with it, in a variety of cloud scenarios to help customers achieve these benefits.  More to come.

9 Responses to The Stack Market: Vblock, FlexPod, VSPEX (and me)

  1. Interesting post. Ive just started looking at converged infrastructure for an upcoming project. Where do you see Oracles ExaLogic and HPs Cloudsystem fitting in, if at all?

    • BlueShiftBlog says:

      Thanks for posting Edward.  
      HP’s Cloudsystem strikes me as a reference architecture, along the lines of FlexPod and VSPEX, but of course it prefers to use HP storage, blades and networking where possible.  If you want an HP flavored solution, it’s seems like a viable reference architecture.  
      Oracle’s Exalogic is a bit unique (and not necessarily in a good way in my opinion).  It strikes me as being a bit inflexible and likely that you may end up buying more hardware than you really need.  See and for more details.

      • jhenry_irl says:

         @BlueShiftBlog Nice blog Kevin. Disclaimer I work for HP in CloudSystem team. HP CloudSystem is not just a reference architecture. It is a completely integrated cloud solution with its own product team, its own part numbers and is available as an out of the box, ready to go cloud direct from HP or through our partners. It is completely open in that we support mutliple hypervisors, multi-vendor servers, storage and networking. There are big advantages to using HP servers, storage and networking due to the level of investment we have made in making them work best together but we certainly don’t insist or force customers to do so. For example 3Par storage is the only storage array platform built from the ground up for Cloud environments with embedded multi-tenancy. We have increased the levels of automation and reduced the amount of steps required in carving out storage so it happens faster with fewer people. Cloud computing by its very nature needs to be open and capable of interoperating across various public and private clouds running on multi-vendor platforms and that forms the basis of HP CloudSystem. We already support Cloud bursting (physical and virtual) to Savvis from HP CloudSystem and others will be added this year. I’d also like to point out that while virtualisation is a very important part of a cloud solution; virtualisation does not on it’s own = cloud. It is an enabler of cloud but any cloud solution worth its salt also needs to cater for orchestration of both physical and virtual infrastructures; it needs to include a self service portal for both IT and end users, it needs to include always on capacity planning and automated DR. These are the basics and if it doesn’t have these features it shouldn’t be called a cloud. Finally one other unique thing from CloudSystem is the availability of Cloud Maps – which turn best in class reference architectures from HP and our partners, Microsoft, VMware, Oracle, SAP, F5 and many many more into templates which can be deployed on CloudSystem at the click of a mouse button. We have over a 100 cloud maps available already free of charge on If you fancy seeing some real world demos of Cloud Maps or other @HPCloudSystem features check out

        • BlueShiftBlog says:

           @jhenry_irl  @HPCloudSystem Thanks for posting James!
          Now that you mentioned it, I do recall discussion of HP’s Cloud System as being more of a product than a reference architecture.  I should have verified this before commenting, but you did a much better job than I could have of explaining HP Cloud System!  
          I’ve used HP servers and storage in the past and I’m sure its a compelling solution for many.  Thanks for correction and info!

        • KendrickColeman says:

           @jhenry_irl  @HPCloudSystem VCE’er here. Nice article Kevin.
          jhenry, here are my thoughts, and i do not want to start a post battle, i just want to put my opinion out there. The is the only reason I don’t consider HP CloudSystem to be a true converged architecture is because even though it’s a single order SKU, there is still going to be 300 boxes shipped to your loading dock and a team of HP Professional Services guys that have to put it all together. It’s still a reference architecture IMO until it’s built on standards. Even if you have two different customers ordering the same equipment, can you guarantee that the 10GB cable 2 meters long connects from port 13 of switch A to port 4 of FEX B on both of them? what about the upgrade process? How can i guarantee interoperability between all my components if I can stick any hypervisor or any type of switch in there? i could go on and on, but this article should answer most everything –
          There is no single answer to everything. Every customer has their choice, and it’s all about solving the business need at the end of the day.

        • jhenry_irl says:

           @KendrickColeman That’s simply not true Kendrick. Customers can have their CloudSystem delivered any way they want; up to and including pre-cabled in a rack with servers, storage and networking so it arrives in a single box if they so wish (by the way every HP solution is available this way – it’s called HP Factory Express). If customers want to order CloudSystem without a LAN and SAN so they can hook it up to their existing core LAN and SAN then our pre-installation workshop (included with the CloudSystem) handles these eventualities of physical placement, cabling etc. The important point is customers do have a choice to order CloudSystem without a SAN and we will connect just fine to any industry standard SAN or iSCSI storage array. Same goes for the core LAN; we just flatten the network and reduce the number of hops required by handling most of the traffic inside the CloudSystem especially in a VM environment where most of the traffic is east to west (VM to VM or server to server) and not so much North South or server to client. This allows customers extend the life of their core LAN infarstructure and defer costly investment when it may not actually be required. Obviously I don’t accept your argument on converged infrastructure; we have been selling converged infratsructure solutions starting with HP BladeSystem cClass back in 2006 and are the only vendor who has servers, storage, networking and management software in Gartners magic Quadrants. If I’m not mistaken I think we are the only vendor to allow vCenter admins manage their servers, storage and networking as well. As regards the article you cited – this was written by a VCE employee and I have lots of HP views should you want them. Let me know and I’ll send on the links 😉

  2. Pingback: Cisco live 2012 – Built on FlexPod - WKSB Solutions

  3. Pingback: Cisco Live 2012 – Built on FlexPod | The Virtual Storage Guy

  4. I think that the FlexPod is a great solution — the clear market success combined with EMC’s decision to attempt to emulate this with their own reference architecture (VSPEX) speaks for itself I think.

Leave a Reply

Your email address will not be published. Required fields are marked *