Renderfarm Workflow
Renderfarm Main | Quick Walk-through | Workflow | Frequently Asked Questions | Advanced Topics |
Contents
Introduction
This page describes the outlines of the workflow of using the renderfarm.
Project- and Filemanagement
This is really something quite generic: you should always do proper Project management and use smart naming of your files. Please refer to the File management page for elaborate information on this topic.
The most important requirements are:
- Don't use spaces or special characters in your project name and your filenames. Limit the names to alphanumerical characters and: `- . _`.
- Make sure all your files (especially textures) are properly put inside your project directory
- Use descriptive names for the cameras that will be used for rendering, eg: "interior_hall_01", "facade_east", etc.
And keep a list with an overview which camera is which.
Optimization
When your scene gets bigger and more complex, it's a good idea to do some optimization. Of course you should always keep a copy of your original scene, before you start optimizing your scene for rendering. This proces may be quite destructive.
Some options for optimization are:
- Delete all history of objects where possible. Pay attention when deleting history as you don't want your model to change.
- Delete parts of your scene (objects) that are not necessary for your render. This depends on the camera view.
- Delete all temporary objects (curves, etc).
- Do a File > Optimize Scene Size to delete unused nodes.
- Replace high-detail geometry with low-detail geometry and smart usage of filetextures (using bump, etc).
- Replace detailed parts of your scene that are far away from your camera with low-detail geometry.
Preparation
Set your project in Maya and load the first scene to be rendered. Open the Renderfarm User Interface (UI). Make sure the project name is correct, otherwise press the reload button to load it from the current Project settings. Reminder: projectname, scenename and jobname must not contain spaces or any other special characters.
Next is to select a camera that will be used for rendering. You can even select multiple cameras.
Rendering multiple cameras
If you want to render multiple cameras, there are two options. Each method has benefits and drawbacks.
1. Selecting multiple cameras:
The quickest to configure: just select the cameras you want to render. There are some limitations however: This won't work when you're going to use image slicing (see below).
2. Creating multiple scenes with different cameras set to renderable:
You can also select a different camera in a copy of the scene, creating multiple copies of your scene each with a different camera selected (renderable). As soon as you select a camera in the renderfarm interface, the camera will be set to be renderable and all deselected cameras will be set to be 'not renderable'. So make the required selection and save your scene using a different filename.
Texture Checking
The Renderfarm Interface will check whether all files used in the scene (eg for textures or image planes) are in the right place and can be found. If there is a problem with your textures, you can use the TOI Texture Tools to fix the problem.
Publish Project
Publishing will copy all the files in your project needed for rendering to a server of the renderfarm. Once your project is published to the renderfarm, the renderfarm can load your files when you submit a render job. Publishing may take some time with large projects.
When you're using references: Referenced scenes are published automatically.
Submit Render Job settings
Enter a job name (eg: "render01" or "eye_street")
There are two main options for a renderjob: you can render a single image or an sequence of images for an animation.
Single Image
When rendering a single image there are two options:
- Render a single image (one or more cameras) on a single computer
- Render a single image (one camera) on several computers (Slicing)
The second option, using image slicing, has the advantage that you can utilize more computing power (quicker).
But there are some limitations to slicing:
- You can only render a single camera per job. If you want to render multiple cameras, you'll have to submit them as seperate jobs.
- Slicing may produce 'banding', where the colors on the edges of two adjacent slices are slightly off. It depends on the render method and settings. Post processing effects such as glow are known to cause heavy banding.
Image Sequence (animation)
When rendering an animation, use this option. You may select multiple cameras. Every frame will be rendered from the selected cameras.
Packet size
The entire job is splitted into chunks of the set packet size. That means that the job is divided into sub-tasks of a number of sequential frames. A small packet size (high number of chunks) will decrease efficiency, as there is a lot of overhead because every node will have to re-read all files of the project when it starts processing a new chunk. A large packet size (small number of chunks) can also be inefficient, because chances are that you won't benefit from optimal parallelization.
Some guidelines:
Animation length | packet size | resulting number of chunks |
---|---|---|
250 frames | 15 - 25 | 10 - 20 |
1500 frames | 50 - 100 | 15 - 30 |
>5000 frames[1] | 100+ | +/- 50 |
Optimization
Be sure to 'bake' your animation: it saves a lot of time. It also may be a good idea to split an animation into several scenes.
Pay attention to optimization when rendering image sequences. A gain of 20 seconds per frame means you'll save up to 6 hours of rendering time in an animation ('movie') of one minute.
Submitting the Render Job
When you press the Submit Render button (either for a single image or an anmiation), the job is submitted to the renderfarm.
Get the results
Refer to Renderfarm Quick Walk-through for information on getting the rendered images from the renderfarm.
- ↑ Try to split large animations into several jobs to keep overview and make it more manageable