5 ms·

Complexity and price are not the problems with Citrix or any VDI solution. The whole concept doesn’t work. I have spent years and years dealing with miserable

by hughrr 5y ago

Complexity and price are not the problems with Citrix or any VDI solution. The whole concept doesn’t work. I have spent years and years dealing with miserable staff who have to sit on the end of a mucked up, crawling semi paralysed Citrix deployment because someone ticked a corporate IT box and got a pat on the back of it. They literally have their potential productivity halved on a good day. Not to mention the pain of trying to get some desktop software I’ve pushed out to standard windows machines fine working with some fucked up implementation of partial profile redirection that some MSP thought was a fantastic idea. Really 99% of Citrix and terminal installs I see these days are box ticking and a false promise of cost savings through density and centralisation. The end story is an end user trying to use a LOB app in chrome with about 200Mb of RAM at their disposal with the CPU and IO rammed at 100% while the MSP who sold them this pile sells them the lie that it’s the LOB vendor who’s at fault. Edit: AWS Workspaces is just as bad. It’s like working with oven gloves on after a heavy night drinking.

rcarmo 5y ago
All the HTML5/canvas clients are pretty much stopgaps for light use. Only a native client will give you a decent experience, since ICA/RDP etc. needs proper hardware access to accelerate rendering, and a canvas in Chrome just isn’t it.
- hughrr 5y ago
  yep agree. But why deploy LOB SaaS browser applications on Citrix? It doesn’t add up.
  - rcarmo 5y ago
    It’s typically about control and auditing. You can record entire sessions, for instance, and limit the ability to scrape sensitive customer data. The list goes on and on if you have a security mindset.
    - hughrr 5y ago
      I’ve seen that promoted but never implemented or used successfully. As for data exfiltration I’ve worked in the defence sector in IT which has pretty stringent security controls in place. At the same time it’s very easy to do without much effort. The key thing with this technology is box ticking to say you have a policy in place and to avoid liability, not to solve the problem effectively. This is only done through picking ethical, rigorous and professional staff. But that doesn’t scale because they are the 1% which is the real problem. So back to liability box ticking.
- imagine99 5y ago
  I know you are correct but I don't really see why it's gotta (continue to) be like that. If legacy protocols like RDP and PCoIP aren't up to a modern, multi-4k-monitor workstation environment, this field is ripe for disruption. Endpoints with surprisingly small footprints can stream multiple 4k videos and support cloud gaming... a potential sourxe for ideas? I would think that there is a lot more potential that even newer protocols such as Blast Extreme are only beginning to scratch...
  - rcarmo 5y ago
    I think you’re wrong to say RDP isn’t up to scratch. I do _optimize_ it, but it now encompasses modern video and audio codecs with very low latency, and it is certainly good enough for me to view training videos, do full screen browser scrolling at a very good clip, etc. Also, it keeps getting improved-today’s RDP has very little to do with what we had five years ago.
    - imagine99 5y ago
      Note that I didn't say it isn't up to scratch. I was referring to others in this thread who said the protocols were subpar. Although RDP still does not support full-screen use on, say, 2 out of 3 monitors in a multi-monitor setup (the 3rd screen being used locally) without the famous VMware Workstation hack. RDP works well for me personally in a variety of use cases (albeit none of them are graphics intensive) - however, user feedback I have received especially from engineers and designers has shown that they are unhappy with it. For example you cannot (anymore) use discrete graphics in an RDP session (e.g. in certain CAD or 3D software). I think this was under development but got removed, iirc. Other protocols like the one used by Splashtop do support this but in our tests RDP on W2k19 still provided higher fps and a smoother experience than Splashtop (with the downside that some software wouldn't start over RDP because it couldn't directly access the graphics card). Are there any particular optimisations you do to RDP beyond what can be done in the "Experience" and "Display" tabs of mstsc.exe?
jordanbeiber 5y ago
This is not completely true - I’ve implemented and maintained remote setups that have worked brilliant. It’s about using it as a tool where that tool fits. One successful example is a support call-center with 2000 users. The specific role used about 30 apps and they were all packaged and maintained by my team. Using a ci/cd fashion to build os image and apps we could roll out updates to base image and sandboxed applications completely seamless to these users. (Side note: We built an automation that synced the two latest versions of all applications to a locally attached SSD drive on all hosts - app startup was near instant for all users which was fantastic at this point in time) Running the clients inside the datacenter brought the latency to a minimum which greatly improved the performance of specific business critical apps. Managing all parts pertaining to logon properly had the users ready to take calls in 10-15 seconds - a great improvement compared to the previous distributed client/server setup. - Reducing latency between application and backend was a major win. Greatly appreciated by users as we saw massive reduction in latency for specific operations (crappy apps). - Being able to integrate with the ip-telephony call-center software required a windows workstation environment hence the choice of “VDI” - No noticeable os and app updates EVER was greatly appreciated by the users. Even when moving from XP to 7 at one point. - Our ability to roll out massive OS and app updates at will to pilot groups who can roll back with a logoff/login was greatly appreciated by the team Granted, this was 10 years ago and my opinion was always that these business applications should be rewritten - BUT I have yet to see a better managed and performing windows client & app delivery. Citix provisioning services (PVS) coupled with app-v/app streaming was an incredibly progressive way of delivering windows workspaces. During my time at citrix I often raised the fact that PVS for linux coupled with docker was an amazing setup - never got any traction on this though… everyone was way too focused on windows delivery.
- hughrr 5y ago
  Did anyone sit with the users for a few days? I hear the same success stories from the MSPs while simultaneously their users are crying into their coffee on a daily basis. In fact “sorry our systems are slow today” is a normal catchphrase these days. Only high performance call centre outcome I’ve seen in recent years is run by Indian scammers.
e12e 5y ago
I remember Sun Sunray working fine back in the day - I don't think the thin client concept / remote desktop concept is inherently flawed? https://en.m.wikipedia.org/wiki/Sun_Ray https://en.m.wikipedia.org/wiki/Sun_Ray
- hughrr 5y ago
  Oh god no those were horrible if you put one next to windows 2000 and had to pick.
  - e12e 5y ago
    Really? For which use-cases would you be able to pick and choose? Running *nix software under w2k would mean cygwin and some kind of commercial x-server or something along those lines? Maybe the commercial ssh.com ssh client?
imagine99 5y ago
I would argue that price and complexity are exactly what caused the other issues you mentioned. Add the ridiculously glossy, pie-in-the-sky marketing mumbo-jumbo that gets sold to execs and combine it with the fact that it us literally impossible to simply click to download it and spin up a lab of VDI setups (Citrix and other major players, even VMware's latest versions) to tweak and study them extensively 9is what leads to rollouts under pressure that too often go horribly wrong. Microsoft offers free 180-day versions of Windows Enterprise foe download at the click of a button that you can test and destroy or even use for a very realistic PoC. Why can't other companies offer that?

by hughrr 5y ago

rcarmo 5y ago

All the HTML5/canvas clients are pretty much stopgaps for light use. Only a native client will give you a decent experience, since ICA/RDP etc. needs proper hardware access to accelerate rendering, and a canvas in Chrome just isn’t it.

hughrr 5y ago
yep agree. But why deploy LOB SaaS browser applications on Citrix? It doesn’t add up.
- rcarmo 5y ago
  It’s typically about control and auditing. You can record entire sessions, for instance, and limit the ability to scrape sensitive customer data. The list goes on and on if you have a security mindset.
  - hughrr 5y ago
    I’ve seen that promoted but never implemented or used successfully. As for data exfiltration I’ve worked in the defence sector in IT which has pretty stringent security controls in place. At the same time it’s very easy to do without much effort. The key thing with this technology is box ticking to say you have a policy in place and to avoid liability, not to solve the problem effectively. This is only done through picking ethical, rigorous and professional staff. But that doesn’t scale because they are the 1% which is the real problem. So back to liability box ticking.
imagine99 5y ago
I know you are correct but I don't really see why it's gotta (continue to) be like that. If legacy protocols like RDP and PCoIP aren't up to a modern, multi-4k-monitor workstation environment, this field is ripe for disruption. Endpoints with surprisingly small footprints can stream multiple 4k videos and support cloud gaming... a potential sourxe for ideas? I would think that there is a lot more potential that even newer protocols such as Blast Extreme are only beginning to scratch...
- rcarmo 5y ago
  I think you’re wrong to say RDP isn’t up to scratch. I do _optimize_ it, but it now encompasses modern video and audio codecs with very low latency, and it is certainly good enough for me to view training videos, do full screen browser scrolling at a very good clip, etc. Also, it keeps getting improved-today’s RDP has very little to do with what we had five years ago.
  - imagine99 5y ago
    Note that I didn't say it isn't up to scratch. I was referring to others in this thread who said the protocols were subpar. Although RDP still does not support full-screen use on, say, 2 out of 3 monitors in a multi-monitor setup (the 3rd screen being used locally) without the famous VMware Workstation hack. RDP works well for me personally in a variety of use cases (albeit none of them are graphics intensive) - however, user feedback I have received especially from engineers and designers has shown that they are unhappy with it. For example you cannot (anymore) use discrete graphics in an RDP session (e.g. in certain CAD or 3D software). I think this was under development but got removed, iirc. Other protocols like the one used by Splashtop do support this but in our tests RDP on W2k19 still provided higher fps and a smoother experience than Splashtop (with the downside that some software wouldn't start over RDP because it couldn't directly access the graphics card). Are there any particular optimisations you do to RDP beyond what can be done in the "Experience" and "Display" tabs of mstsc.exe?

jordanbeiber 5y ago

This is not completely true - I’ve implemented and maintained remote setups that have worked brilliant. It’s about using it as a tool where that tool fits. One successful example is a support call-center with 2000 users. The specific role used about 30 apps and they were all packaged and maintained by my team. Using a ci/cd fashion to build os image and apps we could roll out updates to base image and sandboxed applications completely seamless to these users. (Side note: We built an automation that synced the two latest versions of all applications to a locally attached SSD drive on all hosts - app startup was near instant for all users which was fantastic at this point in time) Running the clients inside the datacenter brought the latency to a minimum which greatly improved the performance of specific business critical apps. Managing all parts pertaining to logon properly had the users ready to take calls in 10-15 seconds - a great improvement compared to the previous distributed client/server setup. - Reducing latency between application and backend was a major win. Greatly appreciated by users as we saw massive reduction in latency for specific operations (crappy apps). - Being able to integrate with the ip-telephony call-center software required a windows workstation environment hence the choice of “VDI” - No noticeable os and app updates EVER was greatly appreciated by the users. Even when moving from XP to 7 at one point. - Our ability to roll out massive OS and app updates at will to pilot groups who can roll back with a logoff/login was greatly appreciated by the team Granted, this was 10 years ago and my opinion was always that these business applications should be rewritten - BUT I have yet to see a better managed and performing windows client & app delivery. Citix provisioning services (PVS) coupled with app-v/app streaming was an incredibly progressive way of delivering windows workspaces. During my time at citrix I often raised the fact that PVS for linux coupled with docker was an amazing setup - never got any traction on this though… everyone was way too focused on windows delivery.

hughrr 5y ago
Did anyone sit with the users for a few days? I hear the same success stories from the MSPs while simultaneously their users are crying into their coffee on a daily basis. In fact “sorry our systems are slow today” is a normal catchphrase these days. Only high performance call centre outcome I’ve seen in recent years is run by Indian scammers.

e12e 5y ago

I remember Sun Sunray working fine back in the day - I don't think the thin client concept / remote desktop concept is inherently flawed? https://en.m.wikipedia.org/wiki/Sun_Ray https://en.m.wikipedia.org/wiki/Sun_Ray

hughrr 5y ago
Oh god no those were horrible if you put one next to windows 2000 and had to pick.
- e12e 5y ago
  Really? For which use-cases would you be able to pick and choose? Running *nix software under w2k would mean cygwin and some kind of commercial x-server or something along those lines? Maybe the commercial ssh.com ssh client?

imagine99 5y ago

I would argue that price and complexity are exactly what caused the other issues you mentioned. Add the ridiculously glossy, pie-in-the-sky marketing mumbo-jumbo that gets sold to execs and combine it with the fact that it us literally impossible to simply click to download it and spin up a lab of VDI setups (Citrix and other major players, even VMware's latest versions) to tweak and study them extensively 9is what leads to rollouts under pressure that too often go horribly wrong. Microsoft offers free 180-day versions of Windows Enterprise foe download at the click of a button that you can test and destroy or even use for a very realistic PoC. Why can't other companies offer that?