DrSpawner actually taking longer to render?

[morgan@dvontz.com]

This is the first time I've ever had a chance to use a slave machine in rendering. Here's the situation...

I followed all of the instructions in the VRay manual in setting up the DrSpawner on my primary machine and on my company office's main server which is the most powerful computer we have as a small company. Here is the rundown of a small test render I did from the VRay Render Progress dialog:

Beginning render sequence
Preparing renderer...
Preparing scene for rendering.
[RenderView] startCameraTime=0.000000, endCameraTime=0.033333
[RenderView] numCameraTMs=2, numFrames=1, frameSamples=2
Initializing built-in VFB.
[RenderView] startCameraTime=0.000000, endCameraTime=0.033333
[RenderView] numCameraTMs=2, numFrames=1, frameSamples=2
Starting DR
Connected to render host (I.P. address)
Using 1 hosts for distributed rendering.
Subpixel color mapping is on: rendered result may have incorrect brightness.
Preparing camera sampler.
Preparing scene for frame.
Compiling geometry.
Preparing ray server.
Scene transferred to (I.P. address)
Server (I.P. address): Scene loaded; starting render.
Creating and initializing 2 thread(s).
Allocating memory for build data of 12588 faces (352464 bytes, 0.3 MB).
Initializing face build data.
Creating 'done' event.
Starting first thread.
Waiting for thread completion.
Releasing thread resources.
Server (I.P. address): Render completed.
Preparing faces for intersection tests.
SDTree statistics:
Total number of faces stored: 12588
Max tree depth: 48
Average tree depth: 17.0445
Number of tree nodes: 3085
Number of tree faces: 27832
Number of tree leafs: 1192
Average faces/leaf: 23.349
Memory usage: 1.30 MB
Scene bounding box is [-24,-20.25,0]-[24.25,22.5,16.25]
Preparing direct light manager.
Preparing global light manager.
Irradiance sample size is 84 bytes
Photon size is 56 bytes.
Light cache sample size is 120 bytes.
1 interpolation maps registered
Rendering interpolation maps with minRate=-4 and maxRate=-1
Setting up 2 thread(s)
Bitmap file "X:\Morgan\Rhino files\ASGVIS\Materials\HDRI\Aversis App_living.hdr" loaded.
Threads completed
Calling endPass() on irradiance maps
Calling prePassDone() on irradiance maps
Image prepass completed.
Rendering image.
Setting up 2 thread(s)
Threads completed
Waiting for DR to finish
Number of raycasts: 3561810
Clearing global light manager.
Clearing direct light manager.
Clearing ray server.
Clearing geometry.
Clearing camera image sampler.
Clearing camera sampler.
Clearing DMC sampler.
Clearing path sampler.
Clearing color mapper.
Closing DR
Render completed

With my very limited knowledge of all of this it still seems like it was successful in distributing and hence dropping the time it takes for a render. However, it actually increased the amount of time by approximately 4 seconds vs just rendering locally on my one machine. Any ideas? I'm completely self taught (aside from all the fantastic webinars) in vray and rendering so I'm sure there is something very simple I'm missing. Any help anyone could afford would be greatly appreciated. Thanks.

[40th]

Re: DrSpawner actually taking longer to render?

what were the render times?

[aclark]

Re: DrSpawner actually taking longer to render?

There are a lot of factors that can affect distributed rendering. I am curious, how large of an image are you tyring to render (pixels)? Keep in mind that there is a lot of overhead time when sending information back and forth accross a network. Therefore, if it is a simple scene and simple render, it's possible that you will not see the true benefits of what DR has to offer. However, if it is very complex with lots of lighting, materials, etc....I beleive you will see a significant difference in the amount of time it takes to render. Ironically I just tested this, and I found that a complex interior scene rendered approx. 30% faster compared to when I did not use the DR.

Food for thought, hope it helps.

[40th]

Re: DrSpawner actually taking longer to render?

Hi aclark,
That was my point. because if the render time were 20 sec vs 23 sec (simple scene) is a lot different then 40min vs 43min.
Good explanation

[morgan@dvontz.com]

Re: DrSpawner actually taking longer to render?

The render times on the simple one were 34s vs 38s. I tried a much larger and complex model (by my puny machine's standards... Fernando's lava lamp tutorial) which normally takes my machine about 25 minutes. The Render progress dialog shows the connection to the slave/server then shortly after shows it's not even responding therefore confusing me even more. I shut down the render as soon as I saw that. Here's that list:

Beginning render sequence
Preparing renderer...
Preparing scene for rendering.
[RenderView] startCameraTime=0.000000, endCameraTime=0.033333
[RenderView] numCameraTMs=2, numFrames=1, frameSamples=2
Initializing built-in VFB.
[RenderView] startCameraTime=0.000000, endCameraTime=0.033333
[RenderView] numCameraTMs=2, numFrames=1, frameSamples=2
Starting DR
Connected to render host (IP address)
Using 1 hosts for distributed rendering.
Subpixel color mapping is on: rendered result may have incorrect brightness.
Preparing camera sampler.
Preparing scene for frame.
Compiling geometry.
Preparing ray server.
Creating and initializing 2 thread(s).
Allocating memory for build data of 866514 faces (24262392 bytes, 23.1 MB).
Initializing face build data.
Creating 'done' event.
Starting first thread.
Waiting for thread completion.
Scene transferred to (IP address)
Server (IP address): Scene loaded; starting render.
Server (IP address): Render completed.
Releasing thread resources.
Preparing faces for intersection tests.
SDTree statistics:
Total number of faces stored: 866514
Max tree depth: 39
Average tree depth: 24.6088
Number of tree nodes: 138051
Number of tree faces: 1861656
Number of tree leafs: 56886
Average faces/leaf: 32.7261
Memory usage: 51.88 MB
Scene bounding box is [-17.5958,-10.3385,-1.81162]-[19.1462,29.2655,29.6845]
Preparing direct light manager.
Preparing global light manager.
Irradiance sample size is 84 bytes
Photon size is 56 bytes.
Light cache sample size is 120 bytes.
1 interpolation maps registered
Rendering interpolation maps with minRate=-4 and maxRate=-1
Setting up 2 thread(s)
Bitmap file "C:\Documents and Settings\fpedrogo.ASGVIS\My Documents\HDRI\Aversis\Aversis Living_day_LatLog.hdr" failed to load: cannot open file
Bitmap file "C:\Documents and Settings\fpedrogo.ASGVIS\My Documents\Featured Download\Lava Lamp SSS\Blue-Gradient.jpg" failed to load: cannot open file
Ping not received from host (IP address) for 60000 milliseconds, dropping connection.
Render host (IP address) is not responding
Threads completed
Calling endPass() on irradiance maps
Calling prePassDone() on irradiance maps
Image prepass completed.
Rendering image.
Setting up 2 thread(s)
Threads completed
Waiting for DR to finish
Number of raycasts: 3466783
Clearing global light manager.
Clearing direct light manager.
Clearing ray server.
Clearing geometry.
Clearing camera image sampler.
Clearing camera sampler.
Clearing DMC sampler.
Clearing path sampler.
Clearing color mapper.
Closing DR
Render cancelled

If I'm not mistaken the "failed to load" items in the middle are because of those materials not being on the slave? This wouldn't not allow it to ping though would it?

[morgan@dvontz.com]

Re: DrSpawner actually taking longer to render?

After doing a number of different tests and scenarios I went back to my simple model. I noticed that the slave was not responding on the lava lamp file so I added more complex materials to the simple rendering including glass, metal and stone. After doing this, the same "ping not received" error in the render progress dialog box came up. What doesn't make sense to me is that consistently in every one of the scenarios at the beginning of the dialog box it states that the scene has been transferred to the slave and "Scene loaded; starting render." shows up as well. If it's doing this it is obviously in communication with the slave so why all of sudden would it not ping and talk to it? Was the first version of the scene without materials so quick of a render that the slave really never did any work with the scene to begin with? To me this would explain the simple 4 second spike in the time since the Spawner has to load and exit at the end of the rendering.

[aclark]

Re: DrSpawner actually taking longer to render?

I have run into that error message many times. It's important to remember that you need to have ALL of your materials in a shared network location for the DR to work correctly....or at least as it should. I can see in your progress window that several bitmaps are being loaded that are on your C:drive, and I have found that this will complicate DR. DR is basically saying "i have 50 seconds to find this bitmap" and when it doesn't, you may get an error, and the salve will not respond. Also, when i was talking about a complex scene, I am talking a resolution upwards of 3000 and render times closer to 2 hours.

[thomthom]

Re: DrSpawner actually taking longer to render?

Just a note: if you are connected by wireless network you might get into all sort of trouble when rendering. Weak signal or small interruptions might ruin things.

[morgan@dvontz.com]

Re: DrSpawner actually taking longer to render?

Thanks for all the help so far everyone. I've gotten all of the texture maps uploaded onto the shared drive on the server/slave so that everything is shared. I'm working on a hard line, not wireless, so that's not an issue, but I'm still getting a time out from the slave saying...

Ping not received from host (IP address) for 60000 milliseconds, dropping connection.
Render host (IP address) is not responding

...even after getting...

Connected to render host (IP address)
Using 1 hosts for distributed rendering.

...and...

Server (IP address): Scene loaded; starting render.
Server (IP address): Render completed.

...it seems that this is happening when the actual render starts to occur after all of the light caching is completed.

Still stumped. Even though I'm trying to utilize the slave for a small rendering that won't make much of a difference in time, I still would like it to run all the way through. I would like to be able to step up my renderings to much larger and complex images and I know that the error message shouldn't be there in order for that to occur. Any ideas?

[aclark]

Re: DrSpawner actually taking longer to render?

One other suggestion I have is to make sure that you are using the same version of DR Spawner as your VRAY engine. Also, when you receive that error, try closing out and then re-opening the Spawner and try rendering again. If that doesn't work, try changin your DR setup to send all materials to a shared folder on your network. It's basically forcing the slaves to read from one location. If that doesn't work, re-install the spawner entirely. I had a similar issue, and just took it upon myself to ensure I had the same version by uninstalling and re-installing.