The requirement is this:

I need to run legacy (Windows Client/server ) applications on disk-less nodes for a customer.

The customer wants to: a) use bare-minimum configuration systems for end-users (disk-less nodes) b) manage application upgrades centrally (the entire setup will likely run on a LAN but WAN cannot be ruled out at this time. either way administration / upgrade / maintenance should not require attending to each PC individually) c) be able to ensure users can access only the legacy app and not be able to fiddle with the OS settings d) have more clarity on whether the app alone can be sent to the nodes or the entire desktop

I am totally confused with what I've read about PXE, Presentation Server, Terminal Services, XenApp, XenDesktop, Hyper-V.

What fits where? What components are essential? Each of these seems to have some overlap and its not clear to me how I should set this up to achieve the above? Hopefully the questions make sense.

Any help is greatly appreciated. Thanks!


I'll try to stay in the Microsoft camp of technologies to keep it simpler. One of the first questions you should ask is hinted at by Jim B, but let’s assume you’re talking "thin clients" which is what a diskless node is:

Can you use presentation virtualization or desktop virtualization?
In more words: Can you get away with RDS (Remote Desktop Services) aka terminal services/Citrix or do you need the more costly VDI (Virtual Desktops)?

Med-V is out as it’s for managing apps on client hard drives and you’re on thin clients, so to me it's between RDS and VDI.

RDS is the good old terminal server now called a Remote Desktop Services Session Host. You install client apps on the server and users use Remote Desktop (RDP) to connect to a single server (or pool of servers) that runs multiple user logon sessions at the same time. These users are using a Windows Server 2008 R2 desktop, and if an admin looked at task manager they would see all the users .exe's running on one server. There are extra features that makes this user friendly through a web browser, load balanced/FT, and Remote Desktop Gateway which runs it all over SSL port 443 to make it Internet firewall friendly.

If your app has special requirements, and you have users who need their own Client OS to themselves, with their own C: drive, RAM, etc. that's where you can optionally deploy VDI on top of RDS. CAD may be a good example of this. So users are still sitting at thin clients, but when they logon to the remote session, they are actually using one of many Windows 7 client virtual machines running separately on one or more Hyper-V servers. They could still be all sharing the same physical server hardware, but now they have their own "world" to play in. Usually this is only deployed when a Session Host (terminal server) won't fit the bill.

Maybe if you describe the usage some more of the application and users we could help you narrow down the field of choices. RDS is great and rather easy to manage (compared to VDI and other options) if you can get the app to perform well running many times concurrently on a single server through multiple logins.

Citrix is usually only need to run on top of RDS when you go big to thousands of users, but even then the newest Windows Server 2008 R2 RDS can get you there.

  • Diskless doesn't necessarily mean thin. I also wouldn't run VDI on top of RDS (I might do it the other way around however). RemoteFX solves the CAD problem with RDS, and MED-V solves the "clients require their own OS" problem. In a very broad sense, RDS is best when the business requires application density in the infrastructure, VDI is best when there is a need for fluid desktop resources or the decision has been made to invest in server infrastructure rather than desktop infrastructure. – Jim B Mar 2 '11 at 1:02
  • All good comments Jim B, thank you. As I understand, Microsoft's thin client VDI solution uses "Remote Desktop Virtualization Host" which is apart of the RDS role, in order to allow remote users to connect to a pool of virtual Win7 machines. To me that's running VDI on top of RDS platform (not RDS Session Host). Is that correct? – Bret Fisher Mar 2 '11 at 2:32
  • @jimb: the reason i used the term 'diskless node' was to emphasize that i don't want to assume anything except the CPU, RAM and peripherals at each node. – alan-p Mar 2 '11 at 6:37
  • I wish I could share more info, but right now "legacy app" with a high likelihood that its a client/server app with a SQL backend is all I have. So a) How does the remote boot happen? What piece handles that? Is my understanding correct that b) with plain RDS then, I can have diskless nodes remote boot and provide users with individual sessions that can be monitored on the server, and c) each individual will share the desktop with other users? d) All this is also accessible through browser? and thanks both bert, jim for the prompt replies. – alan-p Mar 2 '11 at 6:47
  • Many of these should be seperate questions when fully fleshed out. There is no "remote boot". Microsoft Embedded is the OS on Microsoft thin clients and boots to a desktop that looks like reg windows but has no apps except Remote Desktop and maybe IE. You then need to tell users how to connect via RDP or create some sort of logon automation to have them connect (been about 4 years since I've messed with them). – Bret Fisher Mar 2 '11 at 14:23

The microsoft site explaining all of this is windows enterprise. If you are going to run the app over the wan using a server resource to provide the majority of CPU time for all applications then the section to review is virtual desktop infrastructure. If instead you would like to use the users machine for some tasks but would like to run certain applications from the network (which is not what you are describing but fits into the terms you were mentioning) then take a look at the session virtualization link. Finally, if you want to simply manage applications from a central point and not have to worry about compatability or co-existance issues (usually for legacy apps or "metered" applications, take a look at the med-v link. Most of the terms you are mentioning are related to implementing 1 of those 3 flavors of application virtualization (that term used in its broadest sense).

  • trouble with being a noob is one tends to use broad license with terms :). yes, the idea is to use server resources and leave nothing to the thin client. (oops! need getting used to enter saving edits :). thanks for the pointers, will look at them. – alan-p Mar 2 '11 at 6:12

What your customer wants is totally addressed by "OS streaming".

Check products such as Hp Image Manager, Wyse Streaming Manager, Double-Take Flex, CCBoot.

These products use PXE for network booting and add a "virtual disk driver" so that the clients can boot off a server-based shared virtual disk.

When in a WAN/LAN scenario, the usual implementation is to have a (small) proxy server in each "branch", and have this proxy automatically upgrading the virtual disks it serves (for instance using rsynch/rdiff). Some of the products mentioned above have the "manage remote servers" features.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.