Microsoft hates RDSH. What if they killed it? Hyper-V is the new MultiWin!

While doing research for our upcoming DaaS book, (look for it next week!), I spent some time-too much time, really-buried in Microsoft licensing documents explaining licensing for Windows client SA, SPLA, and Office.

While doing research for our upcoming DaaS book, (look for it next week!), I spent some time—too much time, really—buried in Microsoft licensing documents explaining licensing for Windows client SA, SPLA, and Office. One thing that became clear to me was that Microsoft does not like RDSH. I found reference after reference to why VDI is better than RDSH. Some were passive aggressive back-handed digs, while others were right out there, dissing RDSH's session-based foundations.

I guess this should come as no surprise. Ron Oglesby and I joked about this ten years ago when we went to our first MVP conference. Everything we said to the Terminal Server product group at Microsoft was met with them saying, "Yeah, we agree with you. It will never happen because..." The next eight years of MVP conferences for me were no different. It felt like everyone on the Remote Desktop team was there because they'd pissed someone off. I vividly remember a conversation over beers with a the senior member on that team who said, "Look, you have to understand where this company's roots are. Terminal Server will never be a priority."

So now it's 2014. I haven't seen anything in the past ten years to change my mind. When it comes to rich Windows computing, Microsoft loves Windows client. They want to push DaaS customers to buy SA for their desktops and to use "real" Windows client OSes from their providers rather than RDS sessions. (Heck, look at the price increases of RDS CALs in SPLA. It sure feels like this is Microsoft trying to force the market to Windows client.)

Here's where my crazy mind is going around this. What if Microsoft decided that they're going to kill the RDSH option, meaning you'd have to remote 1-to-1 Windows computers?

Think about it. When Citrix invented MultiWin in the 1990s, x86 hypervisors didn't exist. (MultiWin is the core technology that lets multiple users to log in to the same RDSH server. Citrix created it and licensed it to Microsoft in the late nineties.) Back in those days, MultiWin was literally the only option you had to connect multiple interactive users to the same physical server.

But now we have hypervisors. Now we can run a bunch of VMs on a server and get the same effect of RDSH without MultiWin. Sure, this has been available for the past six or eight years, but the immaturity and performance of hypervisors and storage meant that a bunch of 1-to-1 VMs required more hardware (and cost) than the equivalent RDSH solution. But that's not the case today. Heck, now that we have non-persistent VM image sharing built-in to every hypervisor platform, you could argue that that's the preferred, dare I say "simpler" route.

Think about this from Microsoft's perspective? What possible benefit is there for them to keep developing and supporting RDSH? Hyper-V is the new MultiWin. (I originally thought a benefit would be that Microsoft wouldn't have to pay Citrix to license any of the MultiWin components from Citrix, but Citrix's annual reports from 2007-2012 specifically say that their agreements "do not provide for payments to or from Microsoft.")

Maybe Citrix has been clued into this, which is why XenApp is finally moving over to same platform as XenDesktop?

I'm not suggesting that Microsoft will kill RDS. They'll still have that technology, along with its related components like RemoteFX, to connect in a 1-to-1 way to Windows clients and Windows servers. It looks like Windows 8 (and Windows RT even!) have all the plumbing to be Remote App servers.

But in today's world, is RDSH an anachronism?

Join the conversation


Send me notifications when other members comment.

Please create a username to comment.


I'll take your word for it that Microsoft hates RDSH, but where's the proof that they like VDI? If anything, the whole #FixVDA thing seems to indicate that they don't. I've also yet to see concrete proof that they like DaaS. Lack of SPLA, and no support for multi-tenancy seems to indicate that don't like it either.

I'm hoping things will change for the better under Nadella, but until they do, it seems to me the statement "you have to understand where this company's roots are" is spot on: one user, one PC, one Windows license.


A couple comments here:

1. Why would any company kill a product that brings over $700M a year in revenues WITHOUT doing anything? I bet $$$ per head the RDS Team is by far the most profitable team inside Microsoft (revenues in/size of the team). Not even Office gets close to it given how big the team is.

2. The same technologies that make VDI more affordable also benefit RDS. Have you tried running RDS on ILIO as an example? The list of products/technologies that can make RDS much better is the same in many ways for VDI. With the benefit of the extra density.

3. No matter what, Claudio's Law always kicks in. Windows 8 with Office 2013 still boots on a regular PC the same time Windows 2000 with Office 2000 did 14 years ago. That means even with all the advances a desktop OS virtualized (VDI) will always be tied to massive frameworks, bloated apps and so on. Yes it gets better over time but never at the same rate hardware/etc improves. As per Moore's Law from 2000 to 2014 CPUs got 128 times faster. Is Windows booting 128 times faster? Office loading 128 times faster? .NET apps 128 times faster? That is what I thought. So this whole BS that technology will make VDI/etc faster than light is indeed BS as the OS/Apps/Frameworks will always add more and more bloat.

4. Then we have the licensing. IMHO the reason why Microsoft has something in place that makes running a desktop OS somewhere a major PITA (again licensing wise) is because in a way it threatens the core of its business model that has always been the Wintel OEMs. If less PCs are sold how can they recoup money? By imposing such licenses as they do know lots of people/companies moving away from the desktop into smartphones/tablets/etc still have a damn need to connect to Windows at the end of the day (if that was not true why are we all running Citrix Receiver on phones/tablets, VMware View Client, etc). And we all know that will be the case for MANY years to come. So yes, in Microsoft's head that is indeed a way to get the money they lose due to OEM deals back.

5. Given #4 and #1 I truly believe internally Microsoft does not give a damn crap about RDS, as long as they are VERY profitable (what we know they are). Plus all the ISVs around RDS guarantee that profitability as of today (and has been the case for the past 16 years since Hydra became NT 4 TSE). If they were losing money that would be a totally different story. They make money and a ton of it.

My $0.02…



@Claudio, I agree with everything you wrote. But I wonder... wouldn't this all also be true if Microsoft just pushed single instance VMs instead of RDSH sessions? They would still make all that money?


@Brian I guess the math we would have to do is if everyone using RDSH and buying RDS CALs had to switch to some sort of VDA license, would Microsoft get more or less money?

I am no licensing expert but I believe if you have SA you are covered for VDA correct? And not for RDS CALs, right? So right there you may be reducing the money intake as companies that pay today for RDS CALs would not be paying tomorrow if they switch. So today it is SA + RDS CALs payment vs SA only down the road. Does it make sense?

As the majority of RDS license "consumers" I do believe are corporate customers from this little exercise it seems Microsoft would lose revenue. Of course we are not considering the implications that having a much simplified VDA licensing would have on the overall market (growth on the # of people using VDI, etc). And to sum it up we also have to consider the impact that switching everything from RDSH to VDI would have in terms of scalability. Not everyone would be happy with that for sure right now. But down the road, that could be an option.



I have to strongly disagree on that one, as RDSH is complementary to Hyper-V. No matter how you want to put it, doing multiple sessions on the same operating system will *always* provide a whole lot more flexibility for efficient resource sharing. The type of resource sharing done by Hyper-V happens at a higher-level, and I think we want both.

What about RDSH on Hyper-V? Both coexist very well. As for RemoteApp, there really isn't much of a difference between a regular remote desktop server and a RemoteApp server, it is just an extended feature. What makes a difference is the lack of tools to manage RemoteApp publishing on a Windows Client OS. You can either manually add a registry key to allow any app to be launched, or you can add complete registry entries to configure an application allow list with aliases, icons, etc.

I suspect the RemoteApp server feature is available on Windows Client OSes for the simple reason that when virtualized in Hyper-V, you can configure application publishing for virtualized guests. This can be done with personal virtual desktops or with a VM pool.

Now, I do agree that Microsoft seems to be pushing Hyper-V over RDSH. One clear example of this is USB Redirection which was originally only available only to RemoteFX-enabled Hyper-V guests in 2008 R2. The feature was then made available in non-virtualized RDSH in Windows Server 2012.

However, if Microsoft really wanted to kill RDSH, I think they would really need to change the way they currently provide remote access to their virtualized guests: it may not be called RDSH on client OSes, it's pretty much the same server under the hood except it supports just one session at a time. It would not make much sense to kill what they currently rely on.

To properly kill RDSH, they would need to start doing true host-based remote desktop access, instead of using the virtualized guest's embedded RDP server. Hyper-V does offer limited host-based remote desktop access: the console mode, which works even when you are rebooting a virtual machine. However, regular console access only provides basic input and graphics, nothing else.

What about 2012 R2 Enhanced Session Mode? It's essentially the guest's RDP server accessed over the Hyper-V VMBus instead of the network, and then offered by the Hyper-V host for network access. It's great in the sense that you get the complete RDP feature set, but it's still not replacing RDSH.

One reason why Microsoft may want to push Hyper-V is because they tightly control it. I don't know if you've noticed, but even if there are WMI interfaces offered, there is currently no way a third-party can truly extend core Hyper-V functionality, especially not with the VMBus. RDSH, on the other side, exposes many mature APIs upon which vendors have been building extensions for over a decade.


Pretty cool idea and post.  Moving Multiwin down the stack is useful if and only if the OS overhead can be removed as cost structure (cost, which comes in the forms of traditional consumption compounding overhead, as well as in patch management/rolling upgrade costs, etc)

At the end of the day, the user is interested in driving an app.  Whether that app is delivered by a complex multi-user aware OS or a simple, single-purpose OS is irrelevant to them.  BUT to the operator, who cares about roaming profiles, clashes with virtual registries, etc, this would be a pretty nice collapse to the general app remoting architecture.


don't confuse licensing with technology - that can always change.  RDSH the technology will not replace hyper-V, it is used as the basis for fast user switching and much of the session isolation (think services).  There is no reason why MS can't make RDSH more friendly to the client OS - for example make true multi user Windows Client models (would require new licensing models of course).  I think the bigger question is more one of what do customers want.  In my experience customers don't like that RDSH 1) has a poorly understood boundary of security and performance and 2)that it isn't the client OS so has to be managed differently on different cycles by different teams in IT.  As such they find it easier to manage VDI instances politically and operationally.  The question is more one of does RDSH die because the number of users per server for either model make it that VDI (despite the extra software expense) is worth it because it is as just as scalable, the 'sessions' can be moved around between servers and politically it is easier to manage within the IT dept.  Lastly this is when you think of RDSH from a desktop perspective.  When one thinks of it from a published app perspective does the calculation change?


Interesting take.  In our experience - folks seem really happy with a non-persistent & "desktop-skinned" Windows Server, served 1-to-1 over RDP, refreshed on logout .. of course running on free ESXi 5.5 for the added performance / memory dedupe / density.  RDSH could disappear off the planet for all we care, as long as we can still map the single-user RDP server behind the firewall.


I've been saying this for years. Most arguments here are valid, RDS scales better, cheaper, etc. But AppCompat goes away, so you get zero barrier to entry (this is massively important to get people to move off old operating systems). 99% of apps being compatible still means you have to test every one to discover the 1% that won't work.

I'm of the opinion that scale is much less of an issue than management complexity. You'll never truly be able to manage a Windows Server image doing shared apps under the same image of a Windows Client. The facts are it cost more to manage something than it does to buy more hardware and bite the bullet you get hit with on scale. Plus a lot of techs exist that minimize the scale differences.

Regardless Brian's point is "what if" Microsoft just killed it. Funny, I just met with MS recently and I told them this is exactly what I'd like them to do. We needed WinFrame 20 years ago, we don't today, if the licensing wasn't an issue there would be little reason to justify the complexity of two app delivery systems. Windows client should be used for physical, virtual, and app delivery. Citrix can still capitalize on this because XenDesktop can already publish seamless apps from Windows Client.

IMO if they killed RDSH and #FixVDA we'd have a better easier to manage solution, and VDI would get a lot more traction (considering the RDS market size is pretty large).


Listeq  is one of first suppliers who has implemented VDI technology in its Educational multiseat product as alternative to Microsoft's Multipoint Server solution which until version 2012 was completely driven by RDSH technology.

Using VDi makes isolation possible, so some old memory leaking educational applications will not hang all seats running on the same host. Even a mix of different operating systems on same host are possible.