Recently we’ve seen a refresh of VMware’s marketing campaigns, citing that View is “half the cost of XenDesktop.” Most of the people I’ve run into have just found this claim laughable, but what I don’t find so funny is the attempt by VMware to “prove” this by commissioning a whitepaper from industry analysts Enterprise Management Associates (EMA) on the subject.
The premise is this: define a basic VDI (and VDI only) infrastructure, and compare hardware and license costs using VMware View and Citrix XenDesktop. Seems simple enough on the surface, but just a basic review uncovers some major flaws in the whole argument. Here are some of the big problems that we see:
- Bait-and-switch. How would you like it if a car dealer advertised a trim package with leather, alloy wheels and upgraded stereo for a great price – but when you got to the dealership they told you that the stereo only came with an upgrade to a more costly package? VMware’s proposed infrastructure in the paper includes the use of View Composer (Linked Clones) to reduce cost of storage and management, but that feature is only available in the more costly View Premier – NOT View Enterprise. And yet View Enterprise, even though it doesn’t meet the minimum criteria for the environment, shows up in the license cost comparisons. That edition should be tossed out of consideration.
- XenDesktop VDI edition meets the criteria, and is lower cost than View Premier. Unlike View Enterprise, XenDesktop VDI edition does support storage cost savings, using our Provisioning services technology. AND, XenDesktop VDI edition is available in concurrent user licenses (an omission in the calculations in the paper). Therefore, for the infrastructure described, and for customers that are only considering VDI, you need to really compare XenDesktop VDI ($195/CCU, or $95 per user/device [edit: includes first year Subscription Advantage]) with View Premier ($250/CCU + maintenance [edit: NOT included but mandatory, and omitted from the calculations in the paper]), and using the same levels of concurrency. Do the math – hmmm, how do you get that “half the cost” claim again?!?
- The same costing methodology isn’t being applied to both VMW and Citrix. In this paper, there is an assumption that for 10,000 users, customers should have a 10% overhead (to handle “bursts” up to 11,000 users, for example). So the server hardware cost calculations are supposed to account for this, but in the current paper, ONLY the XenDesktop costs include this 10% overhead! The VMware server costs currently don’t include any sort of overhead, thus making the cost calculations unbalanced. In our discussions with EMA, we discovered that this was an honest oversight that is being corrected – but be aware of this miscalculation! By the way, check out some of the cool new features in XenServer 5.6 than enhance server scalability, like dynamic memory optimization.
- Storage requirements are calculated using different methodologies. In the paper, you can clearly see that the View and XenDesktop environments are accounting for different storage costs. The View configuration only looks at desktop OS image and image delta requirements, while the XenDesktop environment includes user data, as well as server installation storage for the various infrastructure components (which don’t have to use shared network storage at all). Clearly some more granularity is needed so you can draw an apples-to-apples comparison.
- It’s not a real solution scenario. There’s no application delivery, no networking costs, no SSL VPN costs, no WAN optimization, and most importantly, no mention user requirements for performance or broad endpoint device support. When you take all those into account, you’ll likely find that the number and cost of all the third party solutions you need to make View work in the real world is significant – while XenDesktop includes many of these out-of-the-box.
Please note that we’ve spoken directly with EMA about these issues and, to their credit, they are making appropriate corrections and revisions. However, keep in mind that VMware commissioned this paper and basing their entire campaign around the current version – so don’t be surprised if they don’t accept the new revisions that debunk the claims!
Bottom line: if you bother giving up your contact info to get this paper, read it with a critical eye and a mind for what your actual business and end-user requirements are. My experience is that most real world implementations need a lot more flexibility and capabilities than what’s described this paper in order to be successful. If you really want to look at costs, you’ll need to consider server, software licensing, storage, networking for desktop and app delivery (not just VDI). And you’ll need to look at the operational and third-party costs, which can be numerous given the limited feature set of View Premier. VMware can make a narrowly defined, rudimentary solution look “low cost” on paper, but the solution that doesn’t work is the most expensive of all!