Multiple jobs on a render node?

[danio]

Hi all - Our office is currently sharing a 3 node in-house render farm. What we've discovered is that using VRay's built in DR system a node can only do 1 job at a time. I even tried running two 3dsmax.exe instances at once and that didn't fly.

What we'd really like is for the nodes to take as many (or at least two) jobs at once and just throttle the CPU. I'm unsure how Render Node licenses would be used in a situation like this if we were able to achieve it but ideally one machine would only take one license...?

Anyway - I'm sure lots of people have run into this situation. I've looked around the forums and seen mentions of backburner and deadline but I was just curious if there is an ideal solution or a consensus as to the best method to divvy up farm processing power in a farm?

Thanks for any thoughts!

Dan

[dean_dmoo]

Surely doing that will just make everything slower? I just send everything as a dr render to backburner and that allows each job to have max ghz (I have different spec nodes). In bb you can set priority per job.

[Morbid Angel]

On windows parallel processing can be bottle necked a lot, having two frames render at the same time will automatically split the cpu load, but the issue is IO. When one process starts or finishes the frame while another is occupying the cpu, the IO will wait until the cpu becomes available and this can take a long time, as long as the frame is rendering, the other frame which completed will just wait.

When you would use this as DR, meaning it would not be able to load a scene while another render is going on, or it will load it too late to contribute. Same issue with BB on the farm.

So your best bet is get another render node

[padre.ayuso]

It seems to me that unless you get more nodes, you would end up having all jobs finish at the same time, even if you could DV up the workload on the CPUs. 2 renderings DV'd up in one machine, takes 40 minutes. One rendering each 20 minutes, both one after the other = 40 minutes. So even if you could do it, you would not gain speed unless you just buy more nodes, like Dmitry suggests.

[danio]

Hi all - Thanks for the replies. We're not actually a rendering studio, we're an architectural office where we use rendering to do a lot of quick studies, in-house graphics, etc. That means random people rendering at random times and doing quick tests. So it's more critical to not have person B waiting on person A than it is to get absolute top performance. We have VRay for Rhino and Max on all the nodes, and it's actually great because the two don't bump into each other. But now we have more Max users and we're all waiting on the others for access to nodes when using VRay's DR.

Anyway, it doesn't sound like there is a great solution out there but I appreciate all the feedback!

Dan

[padre.ayuso]

Yes there is, go RT. Literally, if you just want it for Feedback, you may just go adding a GTX to each machine, and like americans say, Bob is your Uncle! It will be a very fast feedback and will do you loads of good. Also, you can use the Progressive Mode, that should get you an image right away, and that will give you what you need, I suppose.

[danio]

Yes there is, go RT. Literally, if you just want it for Feedback, you may just go adding a GTX to each machine, and like americans say, Bob is your Uncle! It will be a very fast feedback and will do you loads of good. Also, you can use the Progressive Mode, that should get you an image right away, and that will give you what you need, I suppose.

That's a good point about RT. I wonder if RT CPU with DR would 'use up' the nodes for everyone...

We already run GTX 970's in most machines now, I wonder how RT GPU when using your primary card. I imagine the machine would grind to a halt. I talked them into getting me a Titan X in addition to my 970, so I actually use RT GPU quite a bit but I do more rendering than most at this firm.

[padre.ayuso]

That's a good point about RT. I wonder if RT CPU with DR would 'use up' the nodes for everyone...

We already run GTX 970's in most machines now, I wonder how RT GPU when using your primary card. I imagine the machine would grind to a halt. I talked them into getting me a Titan X in addition to my 970, so I actually use RT GPU quite a bit but I do more rendering than most at this firm.

Great idea. U can test using the 970 for rendering, however your machine will halt. Having the Titan X is a great idea, as a matter of fact I'm requesting to upgrade my machine to a new spec with two of these cards.

[Svetlozar Draganov]

Hi all - Our office is currently sharing a 3 node in-house render farm. What we've discovered is that using VRay's built in DR system a node can only do 1 job at a time. I even tried running two 3dsmax.exe instances at once and that didn't fly.

What we'd really like is for the nodes to take as many (or at least two) jobs at once and just throttle the CPU. I'm unsure how Render Node licenses would be used in a situation like this if we were able to achieve it but ideally one machine would only take one license...?

Anyway - I'm sure lots of people have run into this situation. I've looked around the forums and seen mentions of backburner and deadline but I was just curious if there is an ideal solution or a consensus as to the best method to divvy up farm processing power in a farm?

Thanks for any thoughts!

Dan

It is actually possible to run multiple V-Ray Spawners on different ports on the same machine and submit multiple jobs which will be calculated simultaneously.
Additional licenses are not needed, V-Ray will require only 1 render node license even if multiple renderings run at the same time on the same machine.

Here is a quick tutorial how this could be done.

Run as many V-Ray Spawners as you would like via the following command line:

C:\Program Files\Autodesk\3ds Max 2016>vrayspawner2016.exe -TempDir="c:\temp_dir_1" -port=20204

C:\Program Files\Autodesk\3ds Max 2016>vrayspawner2016.exe -TempDir="c:\temp_dir_2" -port=20205

C:\Program Files\Autodesk\3ds Max 2016>vrayspawner2016.exe -TempDir="c:\temp_dir_3" -port=20206

The above commands are for 3dsMax 2016 but they will work with any 3dsMax version.
Please note that for each V-Ray Spawner instance you will need a separate Temp Directory(must exist) and different Port Number.
Note also that the amount of consumed RAM will increase with each V-Ray Spawner instance.

Then when submitting jobs from workstation machines you need to use the same render slave but with different port number:
workstation1:
workstation2:
workstation3:

Let me know if you have any questions.