Page 1 of 1

Not efficient behaviour in Network Rendering

Posted: Wed Aug 28, 2013 1:25 am
by Ernesto
Unfortunately I am experiencing too much unstability in Network Rendering jobs.
I have an animation that was rendering in progresive mode using SL of 5, 10, 12 intervals.
When the system crashed (I do not know why, but my guess is that there is fatigue, as if there would be a memory leak) there were half of the frames rendered at SL 10 and the other half at SL 5.
I tried to resume the job, and as there were no frames at SL 5, I omited that setting, asking the Network to use two SL steps of 10 and 12. Obviously the first frames were rendered at SL10 and the others at SL 5.
What the system did is to start distributing the files from frame 001 on (heavy MXI and MXS) to the rendernodes, to realize that the files were already rendered at SL 10, so it automatically set it as finished and proceded with the next. The net was overloaded because all the files were being sent to and back all the time.

A more effecient procedure would be if the Network could detect that certain files are at the desired SL already, and avoid wasting time sending them to the nodes to open them, and resend them back to mark it as finished.
That would be a great saving of valuable time!

Ernesto