VDI comes to the Raspberry Pi The Raspberry Pi is now a threat to thin clients. Citrix has been fooling around with the Pi as a desktop virtualisation (VDI) target for a while, even releasing a prototype Citrix Receiver for the little computers. That effort was in early 2014. Citrix has since decided it was inefficient to put a lot of effort into creating …
Then that's the first custom version - 5k of these with a small on board storage module - maybe accessed by some magic (jumper shorting) of the input power cable...
Given that this is clearly a bulk application - Monitors/keyboards with USB hubs used to be common place, I'm sure that 5k of them with a WiFi dongle, or preferably an ethernet port, wouldn't be prohibitively expensive - a Pi Zero (2) (yes I know) would be a great little central piece...
Keyboard with a few ports on the back:
- USB power in
- HDMI out
- USB out marked "Mouse"
- Spare USB out
- Ethernet port
"Congratulations. You've reinvented the 21st century Commodore 64..."
That was roughly the aim... (The exact aim was the BBC micro)
The form factor makes some significant amount of sense, given the power we can pack into tiny computers nowadays.
Heck, a couple of AA battery compartments on the back wouldn't go amiss either...
"The PiZero doesn't have an Ethernet port. You'd have to adapt one or a wireless add-on connection via a USB port. For that cost, you've come up to the price of a Pi2 with way more power, so it hardly seems worth it for a PiZero."
I know it doesn't - and I know it only has one USB port - but with USB OTG hub+Ethernet adaptors available for < $4 I suspect we can still come in at ~$10 for the electronics.
And I'm aware that this will increase the power budget, but the power budget isn't really an issue in most cases - what will be powering the monitor for instance?
I'd really like to see a Pi based machine in a Psion5 case with a modern touch screen.
Licenses for bundled software and video requirements beyond what you get for a drone desktop.
Depending on the technology chosen you have either 2D acceleration as a must or 2D + video streaming acceleration as a must. As a result even the most measly thin client used to come with either a Radeon or an Nvidia IGP onboard. A very small percentage came with Via (ex-S3). That isway more than you could get out of an onboard intel as seen in 815 and 9xx series chipsets.
Only some Atom based clients came with Intel IGPs and frankly, they were so bad that they did not sell. At the same time Intel + Nvidia sold quite well despite the premium for decent video.
On top of that, you usually had a license for some management software and the thin client environment bundled into the cost. This is where the cost comes from.
Otherwise, as far as RPI as a thin client, I would never use one for desktop duties. It may be useful in some vertical industries for specific thin client usage scenarios, but it will s*ck even as a most basic word processor. Being able to do HD video is not indicative - it is a stream, it has no latency requirements and you can easily compress it so it does not require a lot of bandwidth.
Real use is something different - 100Mbit Ethernet really stinks as far as latency is concerned in this day and age. You really need 1G for a thin client to be fit for general purpose (not vertical apps) use. Otherwise the lusers start complaining that there is a significant difference between that and a desktop.
100Mbits/s network, well, I use that at home for video streaming and the like, seems to work OK. Is it really such problem?
The problem really isn't the PHY or MAC, but the bus to which they attach (USB). For me, it's simply a matter of setting expectations correctly, which means that users should not expect the same level of network performance that they would for a motherboard connecting the PHY via PCI or PCI Express.
In my testing, I had no problem playing H.264 video up to 50Mbps. Anything over that level had buffering issues.
"You really need 1G for a thin client to be fit for general purpose (not vertical apps) use."
Really? Tell that to my hundred users running zero clients on 100MB links for the last four years. Full Office suite, financial apps, video, you name it, we run it - all without complaints from users. So I'm not sure where you're getting your info, but it can't be from actual use.
Really? Tell that to my hundred users running zero clients on 100MB links for the last four years.
The fact that your users accept abuse does not mean you should abuse them.
I did some specific tests with x86 based HP and Via clients two year ago - rip the box open, (ab)use the PCI slot, plug in a Gigabit adapter and compare performance. They are clearly constrained by bandwidth nowdays, not by CPU or even video. Even a relatively lousy server can easily sustain 1G feeds this days.
Similarly, the difference between a zero and a proper thin client with 1G especially on a modern 1080p or higher desktop is easy to observe and it is not due to CPU - it is the network. It is less visible with Citrix compared to let's say good old X-windows, but it is visible none the less.
I used a VDI session on a Dell 'zero' client as my daily driver for about 9 months this year. Despite being on 'only' 100Mbit, it was perfectly usable and responsive in the majority of cases, even when driving *four* screens (2 x 900x1680, 1 x 1920x1080, 1 x 1920x1200).
Word was perfectly fine. Excel was perfectly fine. Visio was 'ok', but video playback was poor at anything bigger than 'standard YouTube size(tm)'; I imagine some of that was down to the total lack of GPU acceleration at the server end though.
I would also use a softphone client for the Avaya VoIP system through it, which also worked perfectly well.
No offense to you, fellow commentatortot, but I just got finished doing a gig with this major anus who complained about his stupid PC the entire year, that and everyday at about 10:30am the network would hiccup and it got him EVERY SINGLE TIME. What a dumbass! He "had to have" three flat screens for his "work" but failed to produce on his big project, despite the enormous screen real estate. He was okay at producing some minor code and scripts, and generally didn't fuck up the majority of the time, but I just have to laugh at "IT professionals" who claim they can do work on a single Wyse terminal, but still couldn't get shit done with three fucking flat screen monitors! I gave him my monitor and had no trouble meeting my project deadlines on a single Macbook Pro with a measly 15" screen. Unless you're producing videos, or running the NOC dashboard, or do a whopping great amount of writing, then having more monitors is a waste of everybody's time and money. Stop asking for upgrades you don't need, just because the guy in the other cube is a raging anus who somehow can't get through the day without the most powerful laptop and the most screens and keyboards attached to same.
Fact: upgraded kit does not make a good admin a wizard, it's what you do with almost nothing that makes you the wizard. Also not leaning on google search for each and every one of your lame "solutions" is a sign of a true wizard. Google is where IT shitheads look first
I shouldn't feed the troll, but, Google is where anyone with sense looks first - provided you actually understand and learn from the result. It is a waste of everyone's time to re-invent something that's already been done. It also doesn't help that a lot of documentation/help indexing is so poor that Google is often the best way to search it, rather then rely on the help system..
As to multi monitors, I've used them for years, recently ranging from one 12" 1024x768 TFT on a laptop (obviously when I actually started, you were lucky to get 640x480), to two monitors at work, and five at home (one was really only a serial console). It's perfectly possible to work with one monitor, but there's a definite improvement in productivity with two - especially for comparing/referencing documentation, and programming/debugging.
Beyond two monitors the improvement is substantially less - it's necessary to actually plan how to use the extra monitors, rather than throw windows onto one without thinking about it. Monitoring/multiple VMs can be useful, specific development tasks, or media playback are areas that can benefit.
I used a VDI session on a Dell 'zero' client as my daily driver for about 9 months this year. Despite being on 'only' 100Mbit
I am not saying it is unusable. I am saying it SUCKS BRICKS SIDEWIZE compared to the same on 1Gig. You put the luser on a decent 1G rig he will never look back. So it will become "unusable" then, after the luser has seen the difference. Quotes intended - we know its usable, luser will still b*tch to no end in sight.
Granted, my results may be a bit skewed as I used a very good 1G adapter for a comparison which also had TSO offload (and the thin client OS supporting that). One update (in normal glorified typewriter use) in VDI was ending up fully offloaded. In reality, with the exception of some of the Broadcom MACs which went into the early Intel Atom based thin clients, nearly all 1G NICs in shipping thin clients do not have TSO offload so the effect will not be that drastic. It is clearly noticeable though.
I currently use a citrix solution over a 54Mb/s wireless connection in the office. It's fine.
I also frequently use VNC (running in an XVFB) over IPSec to a data centre on the continent - and guess what, that's fine too.
Sure I wouldn't want to stream video on it, but then again the machine it's running on is somewhat underpowered for that anyway.
For most things I can't tell the difference between 10ms latency and 10 us latency - my ears probably could, 10ms is right in the Hass effect "limit" range, and is a useful rule of thumb when doing audio installation designs.
But on a computer - no chance
Most thin clients work well in an industrial environment, we used to have loads of JackPC's (Until desktop support was outsourced)
Slap the thing in the wall, give it POE and connect a mouse keyboard and monitor, assign a policy to the devices MAC. Much better than going, Eww what on earth is that on the case, why does this vent look like it has a beard, etc...
Do Citrix still have the free for 10 users XenDesktop Express edition going?
I know it's not available at the latest 7.1 (?) release level but is 5.x or 6.x still downloadable? (I'd check but can't recall my Citrix login details!)
Still, for rPi2 users who are thinking of VDI it may still be possible to do the server side for free (if the above holds water)
This post has been deleted by its author
Since you need a micro-SD card to boot up the Raspberry Pi, it does have some storage. Up to the full size of the SD card.
What you could do as an admin - is look the size down and not give out the SU account. But people could plug in USB memory sticks/hard drives without much issue.
It's relatively simple to lock down those ports physically. You can lock a case around the unit preventing access entirely (but you will get a call every time some arse manages to get a cable unplugged), or you can go full-on BOFH and epoxy any unused ports shut.
I've never seriously looked into it, but it seemed that VDI (for Windows) worked out more expensive than fat clients due to the way licencing worked - i.e. you had to licence MS Windows desktop for each client, as well as a server licence and any additional third party costs for their VDI solution.
Even without the thin client hardware I struggled to see the cost benefit, or have I got it wrong (I really haven't looked properly into it)?
Lots of non-financial reasons.
Anywhere you have remote offices handling personal information, or simply where the physical security is potentially an issue - you can deploy dumb terminals and keep everything in your nice, PCI/DPA-Compliant data centre.
Of course you could deploy cheapy fat clients and XenApp rather than full VDI, although people could still fall into bad habits and store stuff locally unless you go to the effort of fully locking the remote devices down (which then tends to reduce productivity and ends with users finding "creative" workflows).
There can be certain benefits for app-licensing, or significant hardware savings for demanding users - there could be scenarios where you have 3D modellers, CAD jockeys or animators, who need a hefty Quadro card plus a Tesla accelerator. But who only really strain their workstations in occasional bursts.
Virtualise say, two or three of them using nVidia's GRID product onto a single beefy machine which can give each of them the power when they need it, but saves the expense of three full-fat workstations, two of which are probably idling along at any given moment...
It's the same argument as virtualising servers. At any given time, not all your servers or users will be maxing out their machine. Indeed many of them will be hardly touching 10%, so throw them onto shared hardware.
Not suitable for every case, and Windows licensing eats into the potential savings more than a *nix ecosystem would, but suitable for many depending on their precise business and workload.
Just want to say you don't need a special build any more. With the Raspeberry Pi 2 and Raspbian Jessie you can just install the ARM HF packages from https://www.citrix.com/downloads/citrix-receiver/linux/receiver-for-linux-1321.html .
This is a good guide to get you going:
http://www.martinrowan.co.uk/2015/08/citrix-receiver-linux-raspberry-pi-2-using-raspbian-wheezy/
I see no reason it shouldn't work with Netscalers.
Hope this helps
What might make this a very interesting development is that support for remote offices might now be simplified to:
Have you tried rebooting it?
OK, take a spare out of the safe, and pop the "defective" computer into the post to me. Meanwhile, I'll pop a new spare into a jiffy bag, and it will be with you tomorrow.
I've been doing just this ever since the original Pi came out. They make excellent thin clients with fairly trivial tweaking to make raspbian boot to a custom desktop with extremely limited menu and which opens remmina automatically (and reboot or poweroff if the desktop is quit.)
Initially used plain XDMCP like older much more expensive thin clients on site then FreeNX (which sadly looks like it's come to a bit of a dead end, more's the pity.)
I've only had one returned in a jiffy bag so far - less than brilliant PSU to blame I think.
Remmina supports multiple network protocols in an integrated and consistent user interface. Currently RDP, VNC, NX, XDMCP and SSH are supported.
However, it does not support Citrix - which is what this article relates to. The problem with Citrix on the Pi is the fact this has been a beta side project for one of the devs, with no support for Netscalers. It's still not clear if this build solves the issue or not....
These latest results show it does work and the combination of a commercial packaging with a pretty generic Linux Receiver makes it all come together. Because Citrix cannot officially support plugins is why they have not been able to release something like this on their own.
I've been using Pi as thin clients for around 3 years, not with Citrix though, I'm running them with Microsoft's RDS, cheaper than running Citrix, Citrix only makes sense if you need an acceptable multimedia experience, which my users don't, they use a CRM and Office only. No issues so far, not even a dead SD card.
You can enclose them in a casing that precludes access to the microSD card, requiring the housing to be disassembled to gain access. Additional security could be created by making the unit self-destruct if taken apart by uninformed individuals or by keeping a small venomous creature inside that would incapacitate any perpetrator.
So cheap, in fact, that Citrix wonders if military users won't mind if they're destroyed – intentionally or unintentionally.
They even considered DEC PDPs consumable, using them to determine the propagation speed of underground nuclear explosions: one down the hole to report anything reportable until it was vapourised, then the topside one switched from logging what was sent from beneath to using TDR to measure how fast the EMP 'ate' the cable. The topside one was fitted with core memory so that if they found the system afterwards, they figured they could fit the board(s?) in another system and read out the data.
The salescritter was rather disappointed that they didn't spring for a maintenance contract.
"The salescritter was rather disappointed that they didn't spring for a maintenance contract."
hmm ..
Reason for RMA request:
a) equipment did not survive predicted temperature range;
b) equipment failed due to unexpected power event causing surge/spike;
c) equipment failed due to dust ingress;
d) equipment failed due to memory failures?
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
