1) Using 3 packages for each szene
Indigo behavior on packing one and only scene.pigs and distributing it to the slaves isn't optimal.
The Problem here is, that one single change in the scene requires to upload the whole scene.pigs with all meshes, textures and envmap included
scene_meshes:
Only the meshes of the scene, get updated if something on the geometry changes.
scene_textures:
Often a big part of the scene.pigs is related to the textures, a own pack for textures could reduce network load on final tuning stages extremely.
env_maps:
Since Env_maps get modified not very often, and exr zipes very bad, a direct transfer of th env_map into a env_maps folder in the "network_cache" folder of Indigo would great. By this my slaves would collect a bunch of env_maps I use mostly and check size on each master start if its exactly the same env_map.
Here some of the benefits:
- for animations for each frame having the same textures send to the clients each time isn't Sparta... its CRAZY!
- Highres EnvMaps can be really BIG, send them once and reuse them... recycling ftw!
- The network load for Internet based Network Renderings would become a lot smaller, an therefore usable again for some Projects
- No more multiple GB of data to clear from slave computers after 1 Day session of network render supported testing (personal record ~12GB!)
- at the moment network cache is used ONLY for resuming purposes
2)Each Slave 1 Frame for animation option
Rendering animations on multiple Systems via network mode is fun! connected via LAN a upload each 30 seconds is possible without drawbacks!
The Problem is that on Internet based Network renderings you set the upload time higher to minimize network load. This is necessary, since most used slaves over Internet are not dedicated to be Indigo Slaves, they are often normal Computers used by Friends or family ^^
Due to longer upload periods, on animations there is a lot of Rendering power lost simply by the fact, that the master reaches halt time or halt spp and kill himself and restarts a new one.
All the computed samples from the slaves since the last upload are lost and wasted!
I know the Scenario is more exotic then point 1, but each slave one frame approach would optimize the usage of such internet based renderings!
If everything sucks, then please allow the slaves to save the scene.pigs only in memory and extrude then to HDD, because atm I keep twice the amount of Data, packed and unpacked on my slave HDD :/
Feel free to drop your comments
