Distributed Rendering / Irr. Map question...

[Roodogg]

Hi guys... bit behind the times. Just getting used to DR with Vray 1.5 (I know, primitive)...

 

for animation, I'm used to rendering out my Irr. map with animation prepass but I always skip frames to save calc time. When rendering those frames, vray log tells me there's an error, but it obviously uses the previous Irr. map. and 'bridges the gap'. Seems to give me good results anyway.

 

Unfortunately, when a slave machine tries to load these maps, it gives errors and then gets excluded from the rendering.

 

Question is, do I have to render every single frame of Irr. Map to utilise DR? Or is it something else.

 

Many thanks in advance for you help!

[Tommy L]

AFAIK there's no time saving in using DR over a frame sequence, so if its causing problems you can just go back to non-DR for your sequence.

[Scott Schroeder]

It sounds more of a settings mis-match than anything. The slaves should be able to see the pre-computed imap as long as it is stored in a network location.

 

But as Tom said, for animations it's probably better to have multiple machines cranking on multiple frames than one machine just doing all the work. Though it all depends on your render times. So if you have 5 machines doing a frame every 10 minutes, DR will only become beneficial if you can crank out 5 frames or more from DR every 10 minutes.

[Tommy L]

I think DR actually becomes slower, certainly in 1.5, because if one bucket hangs then your other machines are idle waiting for the next frame to commence.

[Roodogg]

Sorry, guys. I may not have been clear... I'm not using slaves at all during Irr / LC calculations, I only check DR when I render final images. My question was simply that if there is a map missing for a particular frame, the slaves freak out and get dumped by the DR, whearas the workstation seems to utilise the nearest frames' Irr map.

 

Incidentally, with DR on this animation, a frame is around 22 secs, whereas without its over a minute, so gotta be used!

[Tommy L]

Sorry, guys. I may not have been clear... I'm not using slaves at all during Irr / LC calculations, I only check DR when I render final images. My question was simply that if there is a map missing for a particular frame, the slaves freak out and get dumped by the DR, whearas the workstation seems to utilise the nearest frames' Irr map.

 

Incidentally, with DR on this animation, a frame is around 22 secs, whereas without its over a minute, so gotta be used!

 

well yes, but that doesnt make any difference. The DR makes all the machines work on one frame at a time. The overhead there is when the slaves are waiting for the final buckets to complete, invariably the longest bucket times. So DR is slower overall for rendering your frames.

[Scott Schroeder]

Sorry, guys. I may not have been clear... I'm not using slaves at all during Irr / LC calculations, I only check DR when I render final images. My question was simply that if there is a map missing for a particular frame, the slaves freak out and get dumped by the DR, whearas the workstation seems to utilise the nearest frames' Irr map.

 

Incidentally, with DR on this animation, a frame is around 22 secs, whereas without its over a minute, so gotta be used!

 

Then you have a settings issue going on. DR should be able to find and process the saved imap for any frame. Can you post your settings?

[Tommy L]

Then you have a settings issue going on. DR should be able to find and process the saved imap for any frame. Can you post your settings?

 

Hang on a minute, are you submitting your frame sequence to the farm to all machines with DR switched on? Or are you rendering on one machine using the slaves to give you a boost?

If you are sending to all machines and hoping DR will speed up the frames you'll be disappointed. The DR will fail saying its already busy.

If you are rendering on one machine using the power of the others via DR you are SLOWING your frame sequence due to the afore mentioned bucket lag overhead.

[Roodogg]

My workflow is like this.

 

[DR OFF] Process LC / IM and save all to network path.

[DR ON] Render frame sequence. (Local machine begins rendering, usually two or three frames until DR spawner has loaded on slaves, then on subsequent frames buckets appear from slaves and boost rendering speed)

 

I'm not sure how else it would work?

[Roodogg]

Nb I'm not using BB or anything, simply using machines as a boost, yes. I only notice bucket lag on maybe one frame of animation. (I presume as that machine is still preparing - my workstation does that too upon first render!)

Edited July 22, 2014 by Roodogg

[Tommy L]

Submit frames to slaves via Backburner. Turn off DR. It will be faster and your current problem will go away. It wont be resolved, but you wont have to resolve it.

You can probably use DR on your local machine whilst calculating the LC/IM (I think, I do for stills).

[Roodogg]

I think there a bug on the forum as my responses keep disappearing!

 

Was just saying I've always avoided BB so maybe I should take a look. But I only have a MAX licence for one machine and the installs are only trials on the slaves, so hoping DR will sill work with Vray after they run out! (It said somewhere on a tut they do)

[Tommy L]

I think there a bug on the forum as my responses keep disappearing!

 

Was just saying I've always avoided BB so maybe I should take a look. But I only have a MAX licence for one machine and the installs are only trials on the slaves, so hoping DR will sill work with Vray after they run out! (It said somewhere on a tut they do)

 

The trial versions on the slaves is the correct way to install Max. VRay doesnt fully open the software, it just kind of uses it as a shell. If you open the software on the node to run it, you will start the trial period. However, even after its expired the node will continue to function as a slave as normal, you dont need to worry.

Backburner is super simple as long as your network is reasonably well set up. If DR is working then BB should be a snap.

[Roodogg]

Cheers Tom, you've been a massive help - really great that there's such a time difference so when I'm pounding my head against the screen here at 1am, I can still get advice!

[Roodogg]

Hi Tom. Just a thought on what you said. Dropping the bucket size helps with frame hanging. If you go as low as 8x8px there's none at all.

well yes, but that doesnt make any difference. The DR makes all the machines work on one frame at a time. The overhead there is when the slaves are waiting for the final buckets to complete, invariably the longest bucket times. So DR is slower overall for rendering your frames.