DR vs Backburner for animations

[TomG]

Hi all!

Wanted to ask what are the major differences, and advantages and disadvantages, of using Distributed Rendering vs Backburner for rendering animations. I've pretty much always used Backburner, so that each machine renders its own frame in its entirety, with DR for stills so that all machines can help out on the one image.

In some initial testing (just a short 25 frames out of a test scene), DR seems slower than BB, and I think this is due to sending the scene afresh for every frame with DR, while the scene is sent just once at the beginning of the job when using BB? There'd also be sending the image updates across the network, but I think that would be a much smaller addition to the time?

Would there be any plus points to using DR vs BB? Any way to optimize for using DR?

Thanks!
    Tom

[mraw]

Let's say you got horrible rendertimes of x-hours/frame: then DR would come in handy, because when then last frame(s) render(s), the other idle nodes would join to help.

[TomG]

TY! Very true, nothing sat idle, all machines will always be contributing, and as per frame render time goes up, that would be of increasing benefit. My test scene was for pretty fast per-frame times, just a couple of minutes purely for testing and exploring, so I guess there the transfer of the scene for each frame would show up more (assuming that is what is happening :) )

[Frood]

I personally would never render animations by Corona DR at the moment:

- no adaptivity on slaves, as a result noise limit not usable as stop condition, maybe even different results / different sampling focus depending on the "passes local vs. passes slaves" ratio and the participating slaves + recalc interval setting

- less flexible in load balancing, meaning available render capacity (job switch not easy possible, master always occupied in some way)

- less fail save (standard BB handles this quite good when setup properly)

- and there are still a few problems which occur especially during frame changes (using DR or not) and then you find your master stuck at frame #10 next morning.

The only advantage would be what mraw already wrote or the fact that you may get quicker to your first rendered frame to be sure everything is setup right. But you can do this check independently from the "final" submission anyway.

If not using adaptivity you can btw. also consider using both: submit to BB with another bunch of DR slaves.


Good Luck

[TomG]

Also good information!

I didn't realize that last point was possible - Enable DR in the scene, activate DR Server on some nodes, and submit to BB (note - to a node without DR Server running), then the node rendering from the BB job submission will ask the DR nodes to help out. That can be useful even for stills, as it lets you use the network but without the machine you are working on needing to be tied up with rendering.

Haven't tested what happens if I have two nodes rendering through BB and each asks the DR servers to help out, anyone know what happens in that case?

Overall, for animation I still agree that BB is the way to go though.

[mraw]

Overall, for animation I still agree that BB is the way to go though.

Thankfully I wasn't in the situation to deliver animations with Corona. Thankfully because BB is such an ancient crap. I would like to try RoyalRender with Corona. It's free for up to 4 render-nodes, so this fits to the owner of a single-Corona-license(3nodes)
http://www.royalrender.de/

[TomG]

Yes, I should qualify that as "BB or other network rendering managing tools" :) In the above topics, BB can be replaced with your network rendering management tool of choice for comparison to DR when it comes to animations. Never heard of the RoyalRender, will have a look at it some time, ty, free is definitely an important factor!

[mraw]

Royal Render is on the market since 13 years. It is mostly used by VFX and animation-studios. I guess this might be the reason it is unknown to ArchViz-people. Of course Deadline has the bigger piece of the cake, but IMHO RR is the far better software than Deadline.

[TomG]

I'll have a look into it, ty! BTW, diving into this topic in answer to a customer query about the differences between the 2 approaches for animation.

For myself, I stuck with BB because it was free and simple, and does what small things I needed it to do, but will give RR a look (skipped looking into Deadline as it did more than I ever wanted, and cost money to do all that extra stuff I didn't need!)

[Frood]

I didn't realize that last point was possible - Enable DR in the scene, activate DR Server on some nodes, and submit to BB (note - to a node without DR Server running)

It´s possible to do it like this since v1.1 when I remember correctly. And: of course DrServer can run on that machine. We run DrServer 24/7 on all workstations and slaves. You should just not include that specific node where you submit the BB job to as slave until you want to have 2 instances of Max running each using 50% cpu (what actually happens and works but makes zero or less sense :) And don´t use "search lan" feature either of course in this case.

Haven't tested what happens if I have two nodes rendering through BB and each asks the DR servers to help out, anyone know what happens in that case?

The first get´s the help of the slave, any other fails and works without it until DrServer closes his spawned Max, is ready for new connection and receives another request from any master.

And as for BB being old crap: Correct, but you just have to know your enemy, that´s all :)


Good Luck

[TomG]

Thanks Frood! Good to know those extra details on how it works, all seems quite flexible from how you describe. I will find that handy when rendering stills and want to use DR without using the work machine to render!