ScreamerNet without blocking my Layout

IgorPosavec

New member
Hello Support,

i have a questions, after spending countless hours with the setup (i have made it at the end):
Is there any way to let ScreamerNet render on my network nodes without blocking my Layout? I need to work on the workstation further.
I have looked at the command line setup, and i can not find the command to scan and deliver the job to render farm computers. Actually i need this scene delivery window for nodes in Lightwave as standalone application to isolate it from my man work.

My example:
cd /d l:\11_6\bin\
LWSN -2 -cl:\config\ -dp:\scenes\ p:\scenes\001.lws l:\command\job1 l:\command\ack1 1 1 1

This command tries something, but it is only to start it on my local computer.
Does anyone have an idea how to find and enter the parameter so this 001.lws job goes automatically to the network initialized computers?

thanks a lot in advance!
best regards
 

dsol

Grafiks iz us
I'm assuming you're trying to use the built-in Screamernet Controller inside Layout? If so - no you can't use anything else while it's working. Fortunately there's a plethora of third-party SN controllers out there - I highly recommend using one of them (as they run as separate apps, they don't interfere with you using Lightwave itself)

On windows, there's Butterflynet Render, Amereto (IIRC) and many others (google it)
On Mac, there's Screamernet Controller for OSX (though that's just been announced as deprecated - eurrrghh!!!)
 
Last edited:

IgorPosavec

New member
Hello Dson, Rich,

thank you very much for your feedbacks and help. I think i will go for the one i have found as python script. Two Lightwaves did not work, since in the second Lightwave i don't have "Modler" switch button, after the first one is executing the screamernet.
I will report you how it went :)

ps. It could be a feature request for Newtek - just to cut out the screamernet console they have and let it run as separated thread. I don't think this should be a great problem, and it would upgrade the network rendering feature very much :)

thanks
best regards
 

Danner

Pixel Mage
Run screamernet on the second layout you open instead of the first. You can also modify your shortcut (add -0 at the end of the shortcut command) to force LW not to use the hub, and thus be out of sinc with modeler. Then use that for screamernet only.
 

richhelvey

New member
Hello Dson, Rich,
Two Lightwaves did not work, since in the second Lightwave i don't have "Modler" switch button, after the first one is executing the screamernet.

Use the first one - with the Modeller button as your working layout. The second one without it can be your screamernet controller.
 

dwburman

New member
on the Mac I just made a copy of the LightWave app itself. you can also do that to have one set to always run in 32 bit mode.
 

IgorPosavec

New member
Hello to all, this was all very helpful!
@Danner - i used the Python LWController at the end, since it comes at closest to what i am used to do in productions. It looks ugly, but it works simply perfect! Somehow i got used in the past to Maya/Max network-rendering, board tool, one press button solution without reconfiguration science.

My special wish for the future is Distrubuted (Image) Rendering in Lightwave - this would make LW so awesome!
thanks again to all for help, i am getting slowly warm in LW again...
 

dsol

Grafiks iz us
Network rendering in LW is definitely overdue for an overhaul (for about 10 years now I reckon). It's a pain in the *** to set up and easily broken. Some kind of socket-based replacement would be (I think?) more reliable and also more efficient (lower latency). Please let this be one of the things they do in LW12. Or - if they want some extra cash on top (before LW12) - if the LW3D dev group wanted to make a new cross-platform network rendering solution for LW, with auto setup of nodes (and distribution of content and plugins across them) and support for things like distributed rendering for single images, I would happily stump up some cash now.
 

IgorPosavec

New member
Hello dsol
I share your opinion. Looking how much money we waste here each time to setup and keep network rendering working, i could easily pay this directly to Newtek in order to make working stuff. LW11, with 3 time re-installation (Updates) in last months, cost me meanwhile 8 hours of our Tech-Admin. In money terms, i have paid EUR 500,- if not more for something which comes as one button solution even in Blender or Modo.

What I think that is even more dramatic, is the complete lack of distributed single image rendering: working all day with MRay and Vray, LW has despite its awesome look the slowest render i have seen in production. This manifests mostly as soon as you cross the base line and step into Soft Reflections/Refraction & GI domain. In some cases, despite professional help to setup shaders and renderer settings, we have many times longer render time then with any other renderer. Ok, for good quality i am ready to wait, but in the current "freelancer mode", with once workstation per image is not lucrative.
Giving LW the ability to render a scene in distributed way, spread over few extra machines, would buy us life-time. There is nothing worse then to watch for minutes how the image get refined, and then again, and again, for dozens of times for each test. For example, instead of waiting 20 minutes for a decent preview, i could have it in 5 minutes (Vray, MR etc) and go home, go out or go to sleep earlier. This time advantage is especially crucial when it comes to deadlines and crunches. Not every image is a fun-art of a dining room or a busty girl with a sword where it doesn't mind if it renders the image whole night - we have here deadlines of few days to finish whole advertising video clips.
It can't be that other competitors already move from distributed to cloud-rendering, and LW thinks still in the dimensions of freelancers' micro-cosmos - 1 workstation workflow is enough. :)

Ok, as a "newcomer" i should actually not dare to criticize anything, but i mean it positive: LW has some such great art & production aspects and solutions, much, much, much better then those in current Autodesk products - it hurts to see the current rendering technical limitations.
Nevertheless and ignoring this few "edges", i fell in love LW again and use it meanwhile almost every day - LW rocks, has a great look and feel! Keep up the good work, Newtek!
 

Danner

Pixel Mage
For a screamernet controller I have tried: Butterfly Net Render, it was nice but commercial and a bit confusing. Backburner, yes using a 3d max render controller does work with lightwave, it was semi-dificult to set up. But rendered alot slower than single machine rendering. (loads everything again and computes shadowmaps, for each frame) and I also tried Amleto. That one I do recommend.

Give Amleto a go, Last I needed it, I was using LW 11.0. I hope nothing has broken since, but it was very simple to set up and worked on my first try!
http://virtualcoder.co.uk/amleto/

* It's free
* you can do split image rendering
* built in image viewer

and lots of other nice touches like having it automatically render only missing frames from a sequence.
 

dsol

Grafiks iz us
Bruce Rayne's (now defunct) Renderfarm Commander software for OSX used to support multi-machine single image rendering. And in theory, the render engine itself could probably support breaking up an image into (vertical) segments to distribute. Of course, GI pre-processing takes up a significant amount of time, and that's not something that can be split up yet in the current renderer AFAIK.

There's nothing wrong with pointing out the limitations of the current network render system in Lightwave. *Especially* if you've only started using it recently - since it sounds like you have a good understanding of the alternatives.

For what it's worth, Modo's network rendering is (I've heard) a bit problematic too. It uses zero-conf for connecting nodes, which is (in theory) certainly a lot more modern than the arcane text-file-based system screamernet uses, but I've read a few people tearing their hair out trying to get it to work sometimes.

As for mental ray.... well, I don't know anyone who *likes* mental ray. It sounds like a pig to set up, and getting good results from it is something of an artform, compared to say Arnold or even the LW renderer.
 

IgorPosavec

New member
Hello dsol, all

thanks again for serious feedbacks.
After the 10 days production testing, we have installed this lw pyton LW Contoler and Amleto. I used amleto for the first few days - it was awesome, but as soon as i moved the config files away from the default C:\users\xxx\.NewTek\ to the new global LW folder at L:\ Amleto can not work anymore (it renders the empty frames, and it renders it always on local temp). It seems it expects the main config root at .NewTek, where it was installed. I can't get it work anymore. I have put our programmer to fix it, but after few hours we moved further, since there was not much sense to burn more time on this issue.
The Pyton LW Controler worked on, save one exception (which might have been my bad saved scene) over 30 scenes with 20000 frames in a week on 10 machines without problems. It is ugly like hell, but it is stable like a rock, i think i will stick on it for a while and report you if something happens :)

@dsol, huh, i did not know Modo has such problems too.
About MR, well, if you look in other software forums, you wont find anyone liking Lightwave either -it is "pig to setup", as you can read above. :) Now joke aside, MR is a beast, and if you take a look at CGSociety or Expose Art Books (as an example of who-is-who in cg artwork scene), you will notice MR holds about 70% of all rendered and published images, followed by Vray. I wish LW would deliver such amount of exceptional outputs.
It is difficult (if you are novice), it is sometimes moody (if you are intermediate) and it is mighty tool for Film and CG if you are experienced artist. We are MentalRay production house (about 90% of what we do is MR or programmed for MR Shaders) and once you have learned the "language", you don't need to think much about it. The render is incomparable faster to LW when it comes to Blurry Reflections/Refraction, Motion Blur and FinalGather. This is why i suggested DR above, to compensate the speed.
But it is the same thing i experience with LW at moment. I am using LW for modeling, last 10 years everyday, and i have started using Layout one month ago - for the first time. It is a new language, it is sometimes terrible, the project structure is not flexible, it crashes few times a day etc - but i notice how i get every day better and better. I know now that i dare not to press the button "xyz" because LW will crash. I know that if i move in Scene Editor items and press Undo, it would undo my last edit in the scene and i will lost some settings etc etc. Sometimes it saves the scene, i close Layout and save the objects in Modler, and the whole material settings i did in Layout are lost for some reason - i could cry, i press now save each 5 minutes in both applications to be sure it really saved everything. The same "pigs" as in MentalRay, as long as you don't know how the software works :)
But I am getting better - i have finished after 10 days full CG illustrations in LW and customer could not recognize i have switched the software - he even liked the look much better :)
ps. the discussions about LW-MR renderers are anyway irrational, they differ too much.

thanks
best regards
 
Top Bottom