Greetings Bernie,
I am an art teacher who has deployed 12 of your docking stations with great success. I bought more stations before I was aware of the 14 limit discussed here:
I was hoping to get to 18 next year. I am thinking of another server for next year, as you suggest. Just wondering if BOTH servers can be connected and controlled by one instance of Multpoint manager, as Microsoft implies, and if so, how? Is the connection just a USB cable? Can the primary machine boot or shut down the other server?
I have much better performance when all computers are logged on locally rather than going out over the school’s network via RDP.
Thanks!
Will Allik
Old Lyme, CT
And it’s cool to hear you’ve experienced better performance with USB thin clients vs. network connected (RDP) ones. That’s what we’d expect in most cases, but it’s great to have validation in your environment.
Yes, as you cross over from needing one server to two (because of the 14 station limit on Windows), it does introduce some choices.
The two servers don’t talk via USB, and one server can’t control the other in terms of power, etc. So you’re definitely taking on some extra management in coordinating the two server machines.
The most important part of that is user data – can users log into both servers? And what happens when they save data when they’re logged into one today, and another tomorrow?
In short, you can either keep things simple by assigning users to one server or the other, and allow the two servers and clusters of terminals to be completely separate; or Microsoft has a solution with Active Directory and the WMS Server 2011 Premium SKU that allows user accounts to be accessible across machines (but there’s extra cost and initial setup required).
So 14 thin clients with one Windows Multipoint Server in a room is an important threshold if you want to keep things simple – but there are solutions to scale to multiple servers that Microsoft provides.
Hope that background helps! Just let us know if you have any follow-up questions or any request for us. We’d be happy to help however we can.
Bernie,
I continue to be amazed by your level of support… I am running the Premium SKU, because I had anticipated getting myself up to 18, but Windows documentation is not clear on linking the servers. My computing needs are minimal; (mostly viewing pictures)… I have some experience with Puppy Linux, am actually wondering if I should try to implement Fedora 17. Do you know the theoretical limit on USB clients on that OS? I am using UD-160s.
My classroom may be of interest to other art teachers. All 12 students are currently copying masterpieces from Google Art Project using one HP 8200 Core i3 with 8 gigs of ram.
Thaks again…
Thanks for your kind words! The Fedora 17 solution might be worth trying. What’s cool is you can just take your exact same room as it is now, and just unplug the Windows Server and plug in a Fedora 17 machine (or install on different partitions and multiboot), and quickly switch all the terminals from one to the other – nothing on the terminals or how they’re connected needs to change. All behavior is determined by the software on the main server.
Yes, on Linux the limit on # of terminals is higher - it’s hard limited by USB 2.0’s limit of 127 USB hubs and devices per host controller. Then there’s an arbitrary top limit on the number of framebuffers (FB_MAX, currently 32 http://lxr.linux.no/linux+v3.3.6/incl… ). And then you have a soft limit on scalbility/performance with particular server hardware.
Each UD-160 unit is actually at least 6 devices: usb hub, usb graphics, usb audio, usb network (unused), and what you plug in - mouse, keyboard, and anything else. So the limit is around 20-21 per server (depending on how many powered USB hubs and other things you have in the topology).
For the DC-125, it’s 5 devices, so the limit is around 22-25 per server.
So in theory, you’d be able to connect up to around 20-25 USB terminals per host controller (and most modern machines have two USB 2.0 host controllers internally, mapped to different ports on the machine. For example, the common Intel ICH10 chipset family has two http://en.wikipedia.org/wiki/I/O_Cont… ). So on those machines, you’d be able to get up to 32 per computer on Linux, if you carefully make sure it’s no more than 20-25 on each host controller.
Note that we’ve only just recently updated the DC-125 to be compatible with Fedora 17 (all units currently on sale in the USA are, but not yet in Europe). Whereas the UD-160-A/M has been for a long time.
Do you have the DC-125 or the UD-160-A/M for your thin clients?
Greetings… Well, on a Compaq/HP 8200, Fedora 17 Beta, 8 gig ram, i3 proc., I am getting a log-in screen on all 13 monitors currently connected (1 local direct to graphiocs card and 12 UD-160). I only have mouse and keyboard funtions on 8 stations however (local ps-2 and 7 UD-60). Any ideas for getting any further on this?
Functionality is generally much better than Multipoint 2011, but Multipoint was getting me to 13 on this machine (albeit with crashes).
As you know I’m hoping to get to 19, and wonder if it’s just a hardware issue…
While you’re pondering this, I also need to order a server for next year, and wonder what qualities would best get me to the numbers I need to run Fedora with more UD-160s. I assume that more RAM and an i7 hyper-threading chip would help, but maybe I’m overlooking something… maybe USB 3?
Good to hear that some stations are up! Wow - you’re quick!
Let’s try to figure out why 5 of the UD-160s didn’t have keyboard and mouse come up … the main reason I can think of is differences between how WMS and F17 detect the cluster of USB display/keyboard/mouse. F17 might be more picky.
Fedora requires a plain USB keyboard without any extra functionality like USB hub or audio (which some higher-end keyboards have). These change the USB topology and Fedora can’t recognize them as part of the terminal. Is there any chance that the 5 keyboards that don’t work, have built-in USB hubs?
Daisy-chaining the terminals has a limit of 4 deep at most (applies to all USB situations, both WMS and Linux), but Fedora may be more picky. Is there any pattern to the 5 stations that didn’t come up? Are they at the end of the daisy chain or behind the same hub?
Any other patterns that might set those particular 5 terminals or their keyboards and mice apart?
On the server: That’s great - you’ve got it in terms of specs. USB 3.0 isn’t critical yet, but # of cores, memory speed, and memory size are critical.
A midrange core i5/i7 with 8GB+ of DDR3 RAM is a good match for 10+ terminals. Split the USB 2.0 terminals across the two USB 2.0 host controllers on the system, if possible. Go to a higher end i7 class and 16GB+ RAM for ~20 terminals. The 480Mbps limit per USB 2.0 host controller will be the dominant issue at 20.
FYI: Some news in our latest KickStarter updates today, including links to the latest Fedora builds (which don’t need the SELinux disablement, and have other fixes): http://www.kickstarter.com/projects/1…
Thanks again for posting your results and replying back with additional info!
Well, I played around with this last night and gathered some data. Interestingly, when I unplug the home station (VGA, PS2), I gain an operational Plugable station. Also, my Plugables are deployed in 3 groups of 4, each group with its own USB into the server. When all 12 are plugged in, as I said, I get 12 screens but only 8 working keyboard/mouse combos. When I unplug a cable for 4 of the working stations, the non-functional stations instantly become functional. So for whatever reasons the system is keeping me at 8, and since it’s counting the non-USB home station as one of the 8, I don’t think it’s a USB issue…
Appreciate your interest… I’m hoping my classroom can become a model for low-cost use of computers in art rooms, where we mostly just need to look at pictures. Just need to bump up those numbers!
Thanks so much for this report! If it’s not the other things, it may be a bug in Fedora limiting the number of input devices. We’ll get 9 or 10 stations up here, and see if we can recreate what you’re seeing. I’ll report back with what we find.
It’s exciting to have you putting this to use in this real-world scenario. Thank you!
The reason it appears we can get to that many is 8 mice + 8 keyboards + each audio device has a HID input for volume control + a few other random input devices on the system.
When we hit this, the last station will have graphics come up, but the last lines in /var/log/Xorg.*.log is:
[172902.984] (II) GLX: Initialized DRISWRAST GL provider for screen 0
[172903.030] (II) config/udev: Adding input device USB Optical Mouse (/dev/input/mouse7)
[172903.030] (**) USB Optical Mouse: Applying InputClass “Force Input Devices to Seat”
[172903.030] (II) No input driver specified, ignoring this device.
[172903.030] (II) This device may have been added with another device file.
And if we ls /dev/input/event* we see that all 32 evdev slots are taken up (0-31, there is no 32):
We hadn’t been looking out for this limit on the input side (had been focusing on graphics), so we’ll focus in here and see what we can do to get it lifted. It looks like it will take a kernel rev.
I hope anyone out there capable of the kernel rev can recognize how important this is for the future of Linux multiseat, and for education in general. As a teacher, the ability to drop folders of text or images into a single machine for all students to work from is incredibly useful: the same functions may be available with a roomful of machines connected to our school network, but it would be much slower, harder to manage, and not available without being online, which raises all sorts of other issues with distractions and diversions. I also can’t get approval for an entire lab of computers in an art room, and can’t justify the energy expense/carbon footprint, but I have seen what amazing things kids can do when looking at large, on-screen images of their photographs (I’m attaching a 12th grader’s painting done this way).
For now, might there be a script for disabling audio functions on my system? I don’t use them, and this could theoretically open up a few more dev/input slots, right?
Hi Will - Thanks so much for posting a little on how you’re using it. How cool!
I’m with you on the importance of maximizing terminals / server – not just for cost reasons, but most importantly for ease of system and user data management.
To get the number up on Fedora, the disabling of the audio input devices is a great idea … we’ll look into that. And we have one other idea for a workaround. Will post back here when we have something.
Greetings again Bernie… just checking in to see if any progress has been made boosting numbers by knocking out audio or any other way? I’m getting ready to set up my lab for the school year.
I have looked at the bug report. I saw a couple places where there seemed to be hardcoded limits, making simple scaling fixes hard. There are two other solutions based around Plugable devices and hubs, but they need Fedora systems level hacking, not kernel source hacking. I will pursue them elsewhere. I do not know enough about how fedora multiseat detects Plugable gear and launchs a seat to be clearer now.
Hi Will - Thanks for your patience! I was finally able to figure out a way to disable unused devices, and I’m hopeful this will enable you to get to 16 (or at least 15) on your one server with Fedora 17.
It involves just adding one configuration file to your system.