In the last post of my VDI series, I discussed the Microsoft's VDI licensing. Since VDI is licensed per year, it will get more expensive than traditional desktop licensing in the long run. However, we all know that licensing costs are more or less negligible when it comes to calculating the TCO (Total Costs of Ownership).
- Poll: How reliable are ChatGPT and Bing Chat? - Tue, May 23 2023
- Pip install Boto3 - Thu, Mar 24 2022
- Install Boto3 (AWS SDK for Python) in Visual Studio Code (VS Code) on Windows - Wed, Feb 23 2022
People often think that desktop virtualization, like server virtualization, is a way to reduce costs in IT. I think, this is a misunderstanding, which I will try to clear up in this post. I will also say a few words about the so-called Microsoft VDI tax debate.
VDI hardware costs
The typical thin client argument that your client hardware can be less powerful and, therefore, cheaper doesn't count when it comes to VDI because you need expensive additional server hardware for VDI. Obviously, your server infrastructure needs a lot more horsepower than with session virtualization (Terminal Server) because each user runs a virtual machine with a full-blown operating system on the server.
This is also true if you take into account that virtualization technology allows you to use resources more efficiently. 1GB RAM for a desktop PC is so much cheaper than 1GB RAM for a server that all this resource sharing can't compensate. Of course, this does not only apply to memory but to every little chip or screw in your servers. Servers are expensive. Very!
Nevertheless analysts often claim that you can reduce hardware costs with VDI. However, I have never seen a concrete calculation that would really prove this. Such claims are usually based on interviews with CIOs who introduced VDI. But do you really think that a CIO would admit that costs were raised because of his decision? Excel is a very useful application. Change a number here and a number there, and you get exactly the result you wanted.
VDI and central management
The central management benefits that VDI appears to offer are often overestimated. Since end users need hardware to access their virtual desktops, you still have to manage devices outside your datacenter. And because maintaining a VDI environment is technically even more complex than traditional desktop management, I have serious doubts that you can reduce software management costs just because your desktops are running in the datacenter.
When it comes to central management, it is relatively unimportant if all your desktops are in a central place. Central management means that the admin — not the desktops — is at the center. That said, central desktop management was introduced long ago when we gave up sneaker administration. A desktop is a desktop, virtual or physical. Whether you deploy an OS image through Ethernet or Fiber Channel makes no big difference.
VDI and flexibility
Even Microsoft acknowledges that VDI can't help you reduce costs:
VDI does not reduce desktop costs because it can represent a significant up-front investment in infrastructure, including hardware, software, storage, and network.
Believe me, if a Microsoft marketer admits that you can't reduce costs with a certain Microsoft product, it is really true. So why would you need VDI if you can't reduce costs?
It is the extra flexibility that VDI has to offer, i.e. the ability for end users to access their desktop environment from everywhere, that makes this technology interesting. And of course, flexibility always has its price. Thus VDI can improve productivity in some environments. But this improved productivity comes with higher costs.
The VDI tax
Some rumors are going around that Microsoft only axed VECD because they now have their own VDI products. Thus, VECD for SA customers was considered as a VDI tax to impede rivals such as VMware that entered the VDI market long before Microsoft.
Considering that Microsoft didn't take VDI serious until recently, it is more likely they just saw VECD as an additional way to cash in some license fees. I mean, they did the same thing with Terminal Server, so why not with VDI? The philosophy behind this strategy is, if you want to use Microsoft software in an environment with enhanced capabilities, you should also pay more.
The point is that third-party VDI solutions were competing with Microsoft Terminal Server, which Microsoft always viewed as a form of desktop virtualization. And since you have to pay extra fees if you want to run Windows in a Terminal Server environment, then you also have to pay for VDI. Of course, the part that Microsoft's VDI competitors disliked was that customers had to pay extra licenses for capabilities provided by software that wasn’t developed by Microsoft.
Subscribe to 4sysops newsletter!
But why did Microsoft reduce the costs for desktop virtualization even though they added this new VDI technology to Windows Server 2008 R2? I think, it is simply because now, as Microsoft has VDI products, someone in Redmond has spent a little more time with calculating the overall VDI costs and realized how damn expensive the whole thing is. Thus, reducing at least the licensing costs was the most natural thing to do because otherwise all these new wonderful VDI products would just be non-sellers.
The other thing that VDI does not take into account is the user experience. If you want the rich experience you need a PCoIP capable thin client, those are actually more expensive than a low end desktop in some cases. If you have a solid desktop management strategy (SCCM, Landesk, etc), the ROI is reduced even further.
I guess PCoIP is comparable to RemoteFX. I didn’t want to go to deep into this topic now because I want to try RemoteFX first when Windows Server 2008 SP1 is released. To be honest, I don’t really believe that a rich user experience is possible with VDI. I have been hearing such promises since the early days of Winframe. But let’s just wait and see.
I did some calculations 2 years ago when I wanted to implement VMware View at my workplace, around 130-140 employees. The hardware cost savings were huge, but it all depended on using our current workstations as dumb/smart terminals. The current workstations would last an additional 2 years, replacing them with thin clients at 1/2 or 1/3 of the cost of a normal workstation later.
What stopped me was the lack of a thin client OS for the current workstations. We would probably had ended up with some unmanaged linux distro like Ubuntu but what we wanted was a os that only launched the View client and nothing else. The biggest downer was the manageability of the client VM’s. For VDI to work you have to have some application virtualization solution, Thinapp, App-V. The footprint on the VM for some customization per user must be kept at a minimum.
Next version of View shold solve that with better Thinapp + Virtual Profiles
Elvar, it depends very much on the kind of hardware you choose. Of course, if you just take simple PCs for your servers you can save money. However, considering that you need a lot more fault tolerance, reliability, and reserve capacity you have no other choice than to buy super expensive server hardware. This applies not only to your servers but also to your network infrastructure, UPS, air conditioning etc. And if you do it right, even your payroll costs will rise because you need more high-paid IT pros to make sure to get your servers running again quickly if problems arise. A broken desktop can be replaced by student worker, but for servers you need high-paid specialists.
The point is if you decentralize your IT, you run a much higher risk that your whole IT goes down. To minimize this risks you have to invest a lot in your infrastructure and staff. I have done the calculation for Terminal Server before, and I didn’t find any costs savings. And VDI is even more expensive than session virtualization.
Michael,
For us the infastructure cost would mostly be about storage, but since we got a new san few months ago that wont be a problem. We already have 2 almost empty blade centers so we would just need 2+ extra blades.
If the whole thing would go down then along with the VDI enviroment over 90% of our server enviroment would be down as well. Users wouldnt be able to do anything either way.
Moving the workstations to the datacenter would bring the users even more avilability than they have now.
Elvar, this sounds as if you are trying to find a way to make use of your overcapacity. When I talk about reducing hardware costs I mean something different. 😉
Well yes, it ofcorse the what you haves and what you need can be as different as there are number of people thinking about this 🙂
We are just lucky I guess, 130 people, maybe 40 of them in IT and we are a financial company that is built on IT services. The equipment we have is somewhat ridiculous for a company of this size 🙂
Hi Elvar, I guess the next natural question is are you hiring? lol 😉
OS Streaming reduces VDI TCO :
You can, instead of cloning your desktop VMs, stream the disk to them.
I use to do that with HP Image Manager.
Streaming a 10GB XP image to 1000 VMs would then use only… 10GB of Image space.
You also need a little temporary space (really temporary), let’s say a maximum of 2GB per VM, plus the virtual memory size (from the host perspective, not from the guest perspective). 3 to 4GB of disk space per VM are a good approximation of the upper limit you need to consider when sizing the storage for VDI+OS Streaming.
Note that my VMs are configured not to have any “hard disk drive”. They PXE-boot to HP Image Manager server.
It’s fast and efficient.
And when I need to update the image on all my VMs, I just do it once, on the reference image, and it gets populated to all the VMs as soon as my updated virtual disk image is shared and the guest VMs are rebooted.
As for user data, I use folder redirection and roaming profiles.
Very efficient indeed, I would not use VDI without an OS/Disk streaming solution.