In the CG world, Network Rendering could easily be 3d Prometheus’ blazing gift to humankind. The dark ages of wailing and grinding teeth in a seemingly eternal abyss of waiting for renders to finish are no more. With a few clicks of the mouse, a host of computers can be commanded to render a project (or several) and renders that once would’ve taken days, months or years to finish can be done in hours. Like any Grecian myth, however, this boon is not without a few ironic twists. We could get into specifics, but let’s just say there are cases where network rendering doesn’t work. When this happens, dispatching tasks throughout nodes has to be done locally, which means loading up a scene file, setting frame ranges and executing renders on EACH and EVERY ONE OF THEM.
As you can imagine, this workaround can get pretty nuanced, to say the least.
One of our farmers (aka wranglers), Andrew, once had to deal with a project that wouldn’t work with network rendering, so the scene’s frame range had to be split into separate batches, and manually rendered on our nodes. As you may know, an animated sequence render time may vary based on the content of the scene at a given part of the sequence (imagine a space scene where the camera eventually tilts down into a populated landscape). Since the frames were split into conventional batches of 1-100,101-200 etc., some batches would finish sooner than others, which wasn’t the most efficient way to distribute the workload. Andrew instead rendered the entire range on all the nodes but assigned each of them a step (which renders only every nth frame). In doing this, all nodes shared more or less the same load, and all the frames rendered faster, and at the same time.
If at this point you aren’t bleeding from the nose, have a cookie. You deserve it. If you are, don’t worry about it. When you submit a job at our farm, we do the worrying for you. The moral of the story is this:
Our farmers are well trained, capable and work as a team to in the face of any task, no matter how Herculean.