Post to Twitter |
Comments (6) |
Views (1829) |
My role allows me to speak with many different people (customers, technologists, coworkers, administrators, etc). I've been able to see presentations comparing the different desktop virtualization solutions out there. One of the problems I see is that many of the solutions only focus on one aspect of desktop virtualization, and that is the VDI model.
VDI is only one aspect of the entire desktop virtualization solution. This is a concept that many fail to comprehend. For example, I attended Gartner ITExpo last week and was amazed at how many people I talked to only thought about the VDI scenario (you know VDI, allowing you to have a remote virtual desktop running on a hypervisor in the data center). When I talked to people about the other options, I could see their eyes light up.
If you are reading this and only know about the VDI version, the I suggest you take a look at FlexCast to get a better understanding at all of the different options out there (FYI, even the CIO magazine identifies there is more to desktop virtualization than VDI). But in a nutshell, here's the deal... desktop virtualization includes:
- Hosted shared desktop
- Hosted VM-based VDI desktop
- Hosted blade PCs
- Streamed local desktop
- Virtual Apps to installed desktops
- Local VM-based desktop
I want to focus on the Streamed local desktops scenario. This is the one that really got people's attention at Gartner. Why? Because most organizations do not do a big bang effect of replacing their end point devices. Instead, most have a rolling lifecycle where each year a portion of the endpoints are upgraded and over the course of 3-4 years the entire desktop environment has been upgraded. Once the process completes, it starts over, never ending.
Let's now say you are embarking on a desktop virtualization project. It seems like a waste of resources and money to idle those desktops that are only 1 year old. They are powerful enough to run Windows 7 and the latest applications, so why would we not use the hardware we already have? This is where the streamed local desktop comes in. It uses the same XenDesktop infrastructure, the same OS images, the same application layer and the same personalization layer. The only thing that changed is the hardware layer.
As money always seems to speak louder than words, think about it this way: If you have 3,000 desktops and they are replaced every 3 years on a rolling cycle, that means 1,000 of those desktop are less than 1 year old. If you estimate 50-100 virtual desktops on a hypervisor (XenServer, ESX or Hyper-V) then you need 10-20 fewer physical servers, which is a substantial cost savings (and even greater if you are using a hypervisor that costs money).
So I encourage all of you to not think about the VDI-only solution but instead to look at your environment as a whole. Chances are you will see that VDI-only might work for you, but probably isn't the best way to run your business. Think about it this way... You can create documents in Notepad, but would you really base your business on a solution that only does one thing, or would you use a more complete solution like Microsoft Word that gives you options?
Daniel - Lead Architect - Worldwide Consulting Solutions
- Twitter: @djfeller
- For the latest desktop virtualization information visit the Ask the Architect - Next-Generation site
- Questions - Then Ask The Architect
Comments (6)
Oct 29
Doug Rainbolt says:
Your posting was helpful in emphasizing the wide scope of desktop virtualization...Your posting was helpful in emphasizing the wide scope of desktop virtualization. Your focus on streamed local desktops is interesting. Is it possible to segment by application in this model which applications are streamed and which ones are hosted? My assumption is that everything in this model is streamed. I would also like to better understand your thoughts related to monetary savings. I would assume that with streaming model one would consume some server cycles that hadn't existed previously, yet it doesn't seem to be accounted for.
Oct 30
Daniel Feller says:
Remember, whether you stream, host, install the desktop has no impact on how you...Remember, whether you stream, host, install the desktop has no impact on how you deliver the applications, which can also be streamed, hosted or installed. So in this example, we have streamed local desktops and the applications can either be streamed on top of the desktop image, installed within the streamed desktop image, or hosted on XenApp servers completely separated from the streamed local desktop.
With regards to server cycles, we are removing potential virtual machines from a XenServer, ESX server or Hyper-V server and running the virtual OS on the desktop hardware. So if you already own desktop PCs that are capable, use that hardware instead of spending more money on the backend servers to host the hosted virtual desktops.
Oct 30
Claudio Rodrigues says:
If money is to be considered I am not sure if this argument is valid. Desktops ...If money is to be considered I am not sure if this argument is valid.
Desktops became extremely cheap. Enterprise class hardware to host desktops in the other hand, are not cheap at all. Just on the server side, a robust 4-way box with 16 cores, 32 to 64GB RAM and redundant components will cost at least 30-50 times more than a desktop ($300 a pop). This does not account for the extra costs like SAN storage, SAN connectivity, VECD licenses, etc, etc. Money wise given the very bottom prices on desktop hardware these days, a fat desktop environment cannot be beaten (unfortunately). Do not get me wrong, I do see all the benefits of VDI on all its forms. But as of today, with the current state we have for such technology, it is indeed way more expensive. And before someone jumps in to mention 'long term savings' sorry, in such hard financial times justifying huge upfront costs simply does not fly. It does if we were in different times. Not these days.
Oct 30
Doug Rainbolt says:
I'd like to share my thoughts on Claudio's comment. I agree that rationalizing t...I'd like to share my thoughts on Claudio's comment. I agree that rationalizing these investments on hardware costs, i.e., comparing PC desktop costs to server costs, can be a slippery slope. I have to admit that I'm in learning mode here and don't have all of the answers. It does seem to me that when contemplating configuring a server to execute the Flexcast model and comparing this to a number of desktops, that power consumption deltas are one consideration, assuming that performance concerns are off the table. But, there is an operational consideration worth bearing out. I agree that in today's environment rationalizing IT investments of any kind based on total cost of ownership over so many years is not an easy sell. However, what I'm picking up from my readings and interactions is that among many customers there is an immediate operational concern. A recent ESG survey among 480 IT professionals, 65% of the respondents indicated that they are spending at least 13 hours per client access device for annual maintenance and would consider virtual desktops in some form. What is coming out is that many customers, not all, are finding their PC environments becoming increasingly complex, even to the point that the current IT staffs either lack the skills or the time to keep pace in these environments. In fact, some have expressed an openness to explore an externally hosted virtual desktop solution. Perhaps what we're witnessing is am immediate cry for help- organizations have slashed budgets, yet data continues to grow (largely unstructured) and users expect more and more.
Oct 30
Daniel Feller says:
There is a lot of disagreement over costs associated with desktop virtualization...There is a lot of disagreement over costs associated with desktop virtualization if it is cheaper or more expensive for the business, but you hit on a major reason why desktop virtualization is being looked at so closely: management. The thought of having a single image for desktop OS, 1 image for each application, all brought together dynamically is something that should make anyone who has to manage a desktop enviornment think about twice.
Plus, being able to self-heal the environment with the steamed OS and applications helps reduce troubleshooting time. You can't plan for everything the user will do to the desktop and you will always run into issues that require troubleshooting time. With this model, you reboot and you are good to go.
Oct 30
Stephen Greenberg says:
Daniel, I am glad to hear you saying this. We are finding many real world sce...Daniel,
I am glad to hear you saying this. We are finding many real world scenarios where streaming the OS to existing PC hardware is by far the most cost effective and elegant solution. In some cases, it is a desktop shared image, in others it is actually a WIN FLP that acts as a thin client, leveraging hardware the customer already owns.
Starting a couple of years ago one of my phrases has been "PVS to the dsektop might be a better thin client than a thin client!", I think there are many cases where this should be seriously considered.
Hosted VDI Desktop is getting the mind share, but in the end hosted VDI is by nature expensive, complicated and separates the execution from the endpoint.
PVS is under-rated in this area, I endorse your recommendation here!!!
Add Comment