Some of you may have heard of Windows Multipoint server, and even fewer have been lucky enough to see it in action much less get it working for yourself. There's a reason M$ created the Multipoint systems, but I never did understand what they did to make certain features work, as all they did was import and modify other server platforms.
Before I explain my scenario and the purpose of this post, let's get some background on the advantages I seek, and the quizzical array of problems with the MultiPoint OS.
According to M$, Wiki, and various windows forums, "Multiple stations can be added to a WMS host computer by connecting a single monitor, USB 2.0 hub, keyboard and mouse for each station. Hardware requirements for MultiPoint stations are non-proprietary, and virtually any multi-monitor video card, mouse, keyboard and monitor that is supported on Windows Server 2008 R2 can be used to build a station."
In an ideal application, a "station" is composed of the 3 common components (keyboard, screen, and mouse) all directly connected to a single system with the necessary hardware, having the WMS OS installed to manage and partition resources to each "station". The approach taken by Microsoft and the few companies that followed WMS for the brief time period that it was a promoted solution involves the use of dedicated station devices that operate in a similar fashion to VDI thin clients, but are converged to a USB port rather than ethernet. This approach makes the entire basis of Multipoint pointless, thus being part of the reason for the system's very limited popularity in a growing market of VDI solutions with dedicated graphics technology. From what I've read, there were a few such devices that took advantage of physical graphics ports directly and converged VGA and USB to be delivered to a station across a single line, but other than poorly built proprietary KVM modules there is very little information on these devices-- and in any case their technology was flawed from the start and seem to have quickly become obsolete.
As it stands, the #1 reason NOT TO USE MultiPoint Server is the RAM limitation of 32GB, which is a direct result of the system being lazily constructed as a variant of the SBS and WHS platforms. Seriously, 32GB in 2010/2011 ON A SERVER?!? That's just shallow, even for a company whose only real concern is LICENSING profit among other things.
Let's see, any OS based on W7 takes 2-4GB of RAM just to start, especially with as many as 20 users (as per the maximum "allowed") you may as well say the base takes 6GB RAM to start and maintain BASIC operation. Next, consider the fact that the primary target was SCHOOL USE! Emphasis on "SCHOOL USE" because, well, think about it; on average students are most likely to have open a web browser with several tabs, and maybe a couple documents at the same time; For the sake of argument, let's say each student uses 2GB of RAM (aside from the base OS requirement) to have the aforementioned windows open... now multiply that number by 20 people (the max. user load) and account for system overhead and resource utilization fluctuations, you would need anywhere from 48-64GB of RAM to maintain BASIC OPERATIONS, requiring the use of multiple WMS machines SPECIALLY BUILT WITHIN LIMITS and capable of handling multiple humans and a fair amount of I/O resource partitioning, not to mention the even greater problem of people who need all the resources they can get for video or photo editing (okay, maybe they'd have a dedicated system for that, but this is a thought experiment, so just go with it) and there's simply not enough power left for others to log on without potentially crashing the system-- all of which contributed to the predictable demise of WMS, rendering it completely pointless as an operating system.
The next problem is really a questionable annoyance more than anything, but the fact that M$ had to create WMS 2011 to offset the difference of forking Server 2K8R2 RTM (WMS 2010 base) and 2K8R2 SP1 (WMS 2011 base) does not inspire confidence in the "effort" put forth to create WMS in the first place.
What they did in the 2012 version may as well have put the entire idea of WMS through the shredder...
Not only did they base WMS 2012 on the extremely flawed Server 2012 (Windows 8 fork) OS and ignore the need to upgrade to account for the difference of what was fixed in 2012 R2 (Win 8.1) probably for the same reason as the lousy export of Server 2K8R2, but as far as I know (and please correct me if I'm wrong) they FAILED to remove the 32GB RAM limitation!
Last but not least we observe a downright pathetic attempt to revive the MultiPoint platform as a role service in Server 2016. I'm not even gonna bother trying this one simply because 1. I hate Windows 8/10 and anything of the like, 2. I have no real use for the odd amenities offered by systems newer than W7, and 3. the entire archetype of the server I'm building calls for Server 2K8R2 SP1, and has plenty of room to upgrade IF AND ONLY IF, it becomes necessary.
Offtopic Comment
To answer some questions you might be thinking right now, I originally hate Windows 7 and everything newer, but I do appreciate it's 2008 R2 server variant which expands on the familiar design of Server 2003 by offering more dynamic features for management and overall operation. I am a die hard user of the now cult classic Windows XP system, and THE ONLY REASON I even bother with windows 7 is because it offers an enormous advantage for GPU acceleration of common video and image processing, and especially advanced graphical design software. Server 2008 R2 was more or less windfall in my case-- an incidental find that demonstrated the same feature set of Windows 7 with an expanded console of the XP-based Server 2003 OS. I started to like the ease of use of server 2003 that came with advanced features only a server would have, and when I came to understand the 2008 R2 had the same graphics capabilities of Windows 7, I started to experiment with various possible uses of the system. From this, I began to learn what I actually had, and decided to make use of such... decent software... in a more realistic application to determine the direction of my systems research.
Why am I explaining all this, you ask? Because, my scenario is the need for a hybrid workstation server configuration to have at least 2 people (myself and an associate) able to access and operate in the system locally. For this setup, USB is no problem at all, and I can add a second GPU if necessary, but I don't understand the software aspect of multiple local physical logons. I've heard rumors that it can technically be done with any 2008 or newer server system, but there's almost no information outside of MulltiPoint installations as to how this is possible, and very few people have even tried it. There may have been at least 1 report of a successful mod of this sort, but that was forever ago... GPU integration is paramount in my system, and not even the best hypervisor software is remotely capable (no pun intended) of delivering the grade of experience I need without expensive hardware from 1 of the 2 GPU giants... worst of all, NVidia GRID requires some lousy BS license to function, and AMD is just plain weird.
So tell me people, strictly from a software standpoint, how can I get the multi-"station" feature from WMS to work in a regular 2K8R2 machine? I understand USB splitting, but GPU resource partitioning to have 2 or more different workspaces for more than 1 physical local logon is beyond me. Is it even possible to backport (or in this case sideport) this feature to get it working the way I need? Does anyone know of other software companies that have developed their own program to solve this problem?
