Network Simulation question

Control and manage your RealFlow Nodes
Thomas Schlick
Posts: 178
Joined: Tue Aug 29, 2017 12:35 pm

Re: Network Simulation question

Post by Thomas Schlick » Wed Dec 13, 2017 9:31 am

I've just seen in your screenshots that the "Port" value is set to 8080, but the Job Manager window says "Binding to port: 65454". So please try the following settings under "Preferences":

Manager Location: 192.168.0.138
Port: 65454
Default URL: http://192.168.0.138:8080

This should fix your problems.

Shinya Akasaki
Posts: 17
Joined: Wed Nov 29, 2017 6:14 am

Re: Network Simulation question

Post by Shinya Akasaki » Thu Dec 14, 2017 2:18 am

Thomas Schlick wrote:
Wed Dec 13, 2017 9:31 am
I've just seen in your screenshots that the "Port" value is set to 8080, but the Job Manager window says "Binding to port: 65454". So please try the following settings under "Preferences":

Manager Location: 192.168.0.138
Port: 65454
Default URL: http://192.168.0.138:8080

This should fix your problems.

Thank you!!! It fixed my problem.I cant believe it was just the port mistake.

Can I ask you one last question?
I understood about when particles hits something it will break into more particles to form the slash kind of effect, but is there a way to make it not to break up after x number of particles?(like 5,000,000 would be the max)
Is it K Age/K Speed/K Isolation/K Collision daemons? or is there a option at the emitter to stop break after x numbers of particles are made?

Thomas Schlick
Posts: 178
Joined: Tue Aug 29, 2017 12:35 pm

Re: Network Simulation question

Post by Thomas Schlick » Thu Dec 14, 2017 8:51 am

Yeah! I'm really happy that your network setup is running now.

Of course you can ask me more questions ;)
Unfortunately, there's no limit for stopping the breakup, and k Daemons won't help you here. It's true that they have a split option, but you'll get even more more particles when it's active.

One method I can think of is to increase the collision object's "DY Interact Distance" value to something between 0.15 and 0.25. This way, more particles will be affected by friction, roughness, etc. and you'll get a smoother collision with less droplets.

Shinya Akasaki
Posts: 17
Joined: Wed Nov 29, 2017 6:14 am

Re: Network Simulation question

Post by Shinya Akasaki » Sun Dec 17, 2017 10:04 am

Thomas Schlick wrote:
Thu Dec 14, 2017 8:51 am
Yeah! I'm really happy that your network setup is running now.

Of course you can ask me more questions ;)
Unfortunately, there's no limit for stopping the breakup, and k Daemons won't help you here. It's true that they have a split option, but you'll get even more more particles when it's active.

One method I can think of is to increase the collision object's "DY Interact Distance" value to something between 0.15 and 0.25. This way, more particles will be affected by friction, roughness, etc. and you'll get a smoother collision with less droplets.
I'm down to a point where my particles(Started up 10,000 I think) and because of the mesh I have in the scene, it ends up bouncing everywhere breaking into 16milions+ particles(frame 60), and this cause my ram to go all the way to 50ish GB (My system have 64GB) and this makes 1 frame simulation time upto 5-6hours or more...

How does the particles works in network? Does all PC all simulate the particles or does 1 pc do the particles and other PC use that data to mesh?

Thomas Schlick
Posts: 178
Joined: Tue Aug 29, 2017 12:35 pm

Re: Network Simulation question

Post by Thomas Schlick » Mon Dec 18, 2017 12:06 pm

OK. That was indeed a bit confusing for me, because I simply couldn't imagine the number of particles going up to several millions. Not your fault, i was just thinking into the wrong direction. I thought you're talking about the fluid breaking up into smaller droplets when it's hitting the floor.

Although it's obvious I have to ask to rule out a few things:
- Do you have any k daemons or a filter in your scene whth an active split option?
- Is there a sheeter daemon?
- And does the number of particles also grow when you simulate locally?
- Does it also happen when all daemons - except gravity - have been removed?
How does the particles works in network? Does all PC all simulate the particles or does 1 pc do the particles and other PC use that data to mesh?
A scene with multiple fluids can be simulated on different machines through IDOCs, but the particles will no longer be able to interact. If you want them to interact you don't need IDOCs, but then all particles will be simulated one computer. You can use multiple machines to mesh the particles, but in this case meshing has to be seaprate process: first particles, then meshing.

Shinya Akasaki
Posts: 17
Joined: Wed Nov 29, 2017 6:14 am

Re: Network Simulation question

Post by Shinya Akasaki » Tue Dec 19, 2017 3:16 am

Thomas Schlick wrote:
Mon Dec 18, 2017 12:06 pm
OK. That was indeed a bit confusing for me, because I simply couldn't imagine the number of particles going up to several millions. Not your fault, i was just thinking into the wrong direction. I thought you're talking about the fluid breaking up into smaller droplets when it's hitting the floor.

Although it's obvious I have to ask to rule out a few things:
- Do you have any k daemons or a filter in your scene whth an active split option?
- Is there a sheeter daemon?
- And does the number of particles also grow when you simulate locally?
- Does it also happen when all daemons - except gravity - have been removed?
How does the particles works in network? Does all PC all simulate the particles or does 1 pc do the particles and other PC use that data to mesh?
A scene with multiple fluids can be simulated on different machines through IDOCs, but the particles will no longer be able to interact. If you want them to interact you don't need IDOCs, but then all particles will be simulated one computer. You can use multiple machines to mesh the particles, but in this case meshing has to be seaprate process: first particles, then meshing.
- Do you have any k daemons or a filter in your scene whth an active split option?
I only have crown x4, gravity, K collison, K isoluted, K volume in my scene. I do notice that K isoluted have split option but it is not active.

- Is there a sheeter daemon?
No, I dont have sheeter daemon in my scene, but I do notice that crown daemon have sheeter parameter.
Is that the same as sheeter daemon?

- And does the number of particles also grow when you simulate locally?
Yes, Im sure is some kind of setting somewhere...

- Does it also happen when all daemons - except gravity - have been removed?
No, because my scene is a liquid explosion scene, I used the crown to pump up the liquid to hit the ceilling.
With only gravity in my scene nothing happens. Also I want to mention that my scene is made from standard particles and not DY particles. I tried with DY particles but it keep crashing with both GPU and CPU simulation.
P.S. Im using intel i7 7700 with Geforce 1070 GTX

Thomas Schlick
Posts: 178
Joined: Tue Aug 29, 2017 12:35 pm

Re: Network Simulation question

Post by Thomas Schlick » Tue Dec 19, 2017 9:16 am

It's definitely the crown daemon. As you've seen already this daemon has sheeter parameters and creates new particles to fill holes. You've said that you're working on an explosion scene, so your particles will be highly accelerated and RF tries to fill the gaps between fast moving particles. Then you also have 4 crown daemons and maybe they even do overlap. All this together results in a huge amount of particles, and that's no wonder to me.

You also said that you have crashes with Dyverso. Well, it shouldn't happen, but it doesn't really surprise me as well with this huge amount of particles, esp. with the GPU where you have just 8 GB.

You should try to deactivate the crown daemon after a certain frame to prevent it from creating more and more particles. You can do that with the daemon's "Simulate" parameter by animating its state from active to inactive. Or you animate the "Cavities detection radius" from 1.0 to 0.0, or "Min cavity size" to 0.0 over time. You should also deactivate "Use velocity alignment".

Shinya Akasaki
Posts: 17
Joined: Wed Nov 29, 2017 6:14 am

Re: Network Simulation question

Post by Shinya Akasaki » Tue Dec 19, 2017 9:39 am

Thomas Schlick wrote:
Tue Dec 19, 2017 9:16 am
It's definitely the crown daemon. As you've seen already this daemon has sheeter parameters and creates new particles to fill holes. You've said that you're working on an explosion scene, so your particles will be highly accelerated and RF tries to fill the gaps between fast moving particles. Then you also have 4 crown daemons and maybe they even do overlap. All this together results in a huge amount of particles, and that's no wonder to me.

You also said that you have crashes with Dyverso. Well, it shouldn't happen, but it doesn't really surprise me as well with this huge amount of particles, esp. with the GPU where you have just 8 GB.

You should try to deactivate the crown daemon after a certain frame to prevent it from creating more and more particles. You can do that with the daemon's "Simulate" parameter by animating its state from active to inactive. Or you animate the "Cavities detection radius" from 1.0 to 0.0, or "Min cavity size" to 0.0 over time. You should also deactivate "Use velocity alignment".
Thank you for the advice. I will look into the parameters. You mention that my GPU crashing is not a surprise during it only have 8GB of ram, what is a good simulation GPU for a professional work? or CPU?
Is it something like Xeon with Titan cards?

Thomas Schlick
Posts: 178
Joined: Tue Aug 29, 2017 12:35 pm

Re: Network Simulation question

Post by Thomas Schlick » Tue Dec 19, 2017 11:07 am

You mention that my GPU crashing is not a surprise during it only have 8GB of ram
Sorry, it wasn't meant that way ;) I have a GTX 1080 with 8 GB, and it's good for 12-15 millions of particles. GTX boards are in fact the best solution for Dyverso. They're (relatively) cheap and offer high performance. I just wanted to say that I'm not surprised that RF crashes with 50 millions of particles created through the crown/sheeter daemons. But, in fact RF shouldn't crash, it should simply stop or do something else. Your hardware is absolutely top for RF.

Question is if your scene always crashes with the crown daemon? Even with just 10 K or 100 K particles? If yes, then I wanr to ask for one of your crashing scenes. Maybe we can find the error and debug it. By the way, did you upgrade to RF 10.1.2.0162? It's free and we've fixed a lot of bugs concerning the GPU and some daemons there.

Shinya Akasaki
Posts: 17
Joined: Wed Nov 29, 2017 6:14 am

Re: Network Simulation question

Post by Shinya Akasaki » Thu Dec 21, 2017 12:47 pm

Thomas Schlick wrote:
Tue Dec 19, 2017 11:07 am
You mention that my GPU crashing is not a surprise during it only have 8GB of ram
Sorry, it wasn't meant that way ;) I have a GTX 1080 with 8 GB, and it's good for 12-15 millions of particles. GTX boards are in fact the best solution for Dyverso. They're (relatively) cheap and offer high performance. I just wanted to say that I'm not surprised that RF crashes with 50 millions of particles created through the crown/sheeter daemons. But, in fact RF shouldn't crash, it should simply stop or do something else. Your hardware is absolutely top for RF.

Question is if your scene always crashes with the crown daemon? Even with just 10 K or 100 K particles? If yes, then I wanr to ask for one of your crashing scenes. Maybe we can find the error and debug it. By the way, did you upgrade to RF 10.1.2.0162? It's free and we've fixed a lot of bugs concerning the GPU and some daemons there.
Ah, I see. I have notice even with CUDA checked and my GPU selected my GPU usage still around 1-2%. Is that normal?
And about the crown daemons, if I have just one on, it seems find but very slow compare to with CUDA off. Then if I turn 2 on my simulation becomes kind of unstable and sometimes even crushes if I try to stop the simulation, lastly if I turn all 4 on it will crash at frame 21 everytime. I think it might be the particles overlapping too much like you said before. This doesnt happens if I use standard particles, is there some kind of difference on the simulation with DY and standard ones?

Yes, Im using 10.1.2.0162.

I have PM you my crash DY scene.

Post Reply