[X2Go-User] optimizing x2go resource use

[Jordan Poppenk]

Dear Mihai,

Thanks for your quick reply! It sounds like switching to KDE is likely a good step for saving on resources. As for colour depth, it sounds like it may not have much effect on bandwidth. But how about server CPU usage - is this a demanding part of the server's contribution?

Cheers,
Jordan


Jordan Poppenk, Ph.D.
Canada Research Chair in Cognitive Neuroimaging
Department of Psychology and Centre for Neuroscience Studies
Queen's University
http://popmem.com
613-533-6009












On 2016-02-09, 7:12 AM, "Mihai Moldovan" <ionic at ionic.de> wrote:

>On 09.02.2016 08:41 AM, Jordan Poppenk wrote:
>> 1) Does X2Go do its desktop rendering on the server's GPU? That is, in building a server to support X2Go sessions, should I buy a fancy graphics card when trying run many sessions; or alternatively multiple core CPU's? If GPU matters, does it matter whether the card is of the GeForce vs. Quadro variety?
>
>Currently no, not at all.
>
>
>> 2) Does reducing colour depth have any substantive effect on the CPU/GPU/bandwitdth burden on the server? My users don't really need more than thousands of colours. I know this and JPEG quality can be adjusted client-side, but when I played with these settings, it didn't seem to impact server-side bandwidth (which continued to hover around 1Mbit per client even with just a few colours). Should I be adjusting this in the LXDE startup settings instead?
>
>In theory, yes. A lower color depth normally translates to smaller images in
>size, but of course the general compression ratio achieved by the algorithm -
>which is also based on properties of the raw image data - might make this point
>moot (especially if the image doesn't actually contain a big variety of colors.)
>
>So that question is actually quite difficult to answer definitely, "it depends."
>
>
>> 3) Is it fair to estimate that I need about 1Gb of RAM per client (just to load the desktop) just to load LXDE as the desktop?
>
>1 GB sounds high for LXDE, but I have never measured this. Maybe others did.
>Still, the problem here, too, is that the answer is: "it depends." For instance,
>consider Java-based applications: these use separate VMs for each invokations
>and do not share any internal state, so the individual footprint is high.
>Compare that to, for instance, KDE(4), which as far as I know does a decent job
>of sharing libraries - hence only the first started session will have a huge
>impact, with subsequent sessions adding less to the global memory footprint (not
>"nothing", of course, because only libraries are shared, no actual program data.)
>
>
>
>Mihai
>