LuxRender network communication is SLOW

[cwichura]

I suppose a creative optimization would be to automatically disable compressing the film file when writing it to disc if the master is detected as being on the same local network as the slave. This would reduce the time spent under lock during DoTransmitFile(). But I suppose it'd be a lot simpler to just add a --compression-level commandline option to luxconsole... heh.

[SATtva]

So, how about we make -W default, and then expose the option to write directly to disk (ie not via a memory buffer).

Lord, if you're going to go this way, i beg you to introduce an option to not write film to disk in slave mode. I'm running slaves on diskless nodes so any disk writes are to be avoided as much as possible, and i'm pretty sure i am not the only one with this setup.

btw, the default of 180 seconds for slave fetches is WAY too frequent. At a bare minimum, it should be like 900 and more realistically, 1800.

+1

[Lord Crc]

Lord, if you're going to go this way, i beg you to introduce an option to not write film to disk in slave mode. I'm running slaves on diskless nodes so any disk writes are to be avoided as much as possible, and i'm pretty sure i am not the only one with this setup.

Oh absolutely, my thought was to essentially invert the -W option behavior.