Setting up a network render

Setting up your own render farm can be done in a few simple steps:

Step 1

Make sure all your computers are properly connected to the network, and check the read/ write permissions on each machine, for example by checking that the currently logged-in user can write a file in the folder you intend to specify as the output folder for network rendering. Make sure there are no firewall restrictions on the machines prior to initializing the Maxwell Render network.

Step 2

Launch the Manager.exe on the computer that is going to run as Manager, and launch RenderNode.exe in the computers that are going to work as nodes. Remember that you can also run a node on the computer that is running the Manager. It can manage the network rendering and also contribute to the rendering process.

Step 3 

Launch the Monitor.exe on the computer where you want to control (launch/ display/ stop) the rendering process. The Monitor will automatically connect to the network to search for the Manager and the available Nodes. They will be listed in the Nodes panel. The Monitor can run on the same computer that is running the Manager. Check in the Monitor>Nodes panel that all your computer nodes are listed there. If some of them are not listed, check the firewall settings on that computer and make sure it is not being blocked.

Step 4

Use the Add button to submit a job. This will open the Add Job Wizard to help you through the process. Select one of the following options.

• Single job: to make a non-cooperative render queue with each node working on an independent frame.
• Cooperative job: several computers work on the same image, which will be merged at the end of the process.
• Animation job: select a scene and the frame range, and the frames will be distributed among the available nodes. Remember that you can also render single frames from the same sequence by using a semi-colon ";" when typing. For example: 1-10;12;20-23 This will render frames 1 through 10, frame 12, and finally frames 20 through 23. Any padding number is allowed.
• Batch job: This option allows you to select multiple scenes to be rendered. All the scenes are rendered using their own settings and output paths. You can perform changes in the rendering settings (render time, SL, resolution, camera, etc), and this changes are applied to all the scenes added. An interesting feature is that in this mode you can choose the Batch Type, between Cooperative type (all render nodes work together in the same scene before starting with another) or Single type (each render node renders a scene separately).

Step 5

Select the networked path of the scene to the output files. Set the frame range (for animations), and set the Render Options (Time, Sampling Level). You should also specify an MXI output, not just the image output. It is possible to select bit depth and file format independently for each render channel. They can be saved in the same folder. The "send dependencies" option will send all files the scene is dependant on (textures, IOR, HDR files) to the Render Nodes. This is very useful to avoid problems when too many computers are trying to read files from the same folder. You should leave this option checked unless you have a very fast file server. If you wish to restart a network render, the Maxwell network will check if the files are already on the Render Node and will not transfer them again.
You can visualize the progress of the network render in the Jobs window in the Monitor. You can stop the process, remove any job, add more jobs to the queue, and display the images that are being computed by the network.

Choose a networked path to locate the source files of the project (scenes, textures, materials, etc). It is highly recommended to use networked paths for your project files, especially for the output files.

Although highly recommended, it is not necessary to also put the scene in a networked path; it is enough if the Monitor can access it through the open file dialog. In addition, if the "Send Dependencies" checkbox is on (default), the scene-dependent materials do not need to
be in networked paths, as they are sent from the Monitor to the Rendernodes. This option is faster and has several advantages, but it requires that the Monitor that submitted the job remains on until all nodes start.
However, if the "Send Dependencies" checkbox is off, the scene-dependent materials should be located in a networked path accessible from all the Render Nodes, that will pick them all as they need them. The output image paths should be network paths accessible from the Manager, which is the machine that is going to save the final output there.
This option ("Send Dependencies" disabled) produces a slower file transfer, but it can be useful when you need to switch off the Monitor that submitted the job, or when you have a fast repository server for scenes and textures.
You can visualize the progress of the network render in the Jobs window in the Monitor. You can stop the process, remove any job, add more jobs to the queue, and display the images that are being computed by the network.

Step 6

Choose the computers you want to work in this job.

You can also check the "Use any render node available" checkbox, and the Monitor will use any node in your farm as soon as they get available.

Step 7

Resuming a network job: Resuming jobs in the Network System is possible for Single, Cooperative and Animation jobs.
When submitting a job, the system checks automatically if the MXI output exists, and if so it asks you to resume or start the job again.

First make some simple tests in your network before attempting a real project or final rendering to ensure that all the Render Nodes can connect and the output is written properly.