Building a hybrid render farm on GCP—new guide available
Senior Cloud Solutions Architect
Visual effects and animation houses in the media and entertainment industry all tell us the same thing: They want to "get out of the hardware business.” And they're right! Purchasing, maintaining, monitoring, and upgrading hundreds (or thousands) of nodes in a render farm makes no sense if you're in the content creation business. Render farms are expensive to build, time-consuming to maintain, and, depending on your location, require a lot of electricity to power and cool so many nodes.
Many visual effects and animation companies have invested millions of dollars in on-premises render farms that run 24/7/365. These always-on resources cost money no matter what's rendering or, more shockingly, what's not rendering; idle resources are typically kept running because power-cycling individual nodes can shorten the hardware's lifespan. And when the workload starts to rise, render farms generate more heat, meaning higher HVAC and electricity costs.
Worst of all, your most valuable resource—your artists—are sitting idle, costing you something that’s worth a lot: the possibility of creating something truly amazing.
An on-premises render farm is also a finite resource, limited not only by the number of CPUs that can be installed in a given space, but also HVAC, power, network, file system performance, and floor load. Factor in the cost of acquisition, depreciation, upgrades, and a team of admins to maintain everything, and the cost of ownership becomes prohibitively expensive to all but the biggest facilities.
Not only that, the natural capacity limitations of an on-prem render farm, no matter how large, restrict a studio’s ability to take on more projects and say “yes” to clients.
What if you weren’t limited to your local render farm? What if you could seamlessly render on cloud-based virtual machines (VMs) using your own render queue manager? Cloud VMs have the distinct advantage of high availability and elasticity to accommodate any size job. And when the job is finished, you (or your render queue manager) can stop the VM, so you only pay for the compute time used.
Google Cloud Platform (GCP) allows you to expand your render capacity by thousands (or hundreds of thousands) of CPUs. Run lightweight compositing jobs on 2- or 4-core VMs. Throw your heaviest, most memory-intensive renders at a 160-core VM with up to 3844 GB of RAM. Attach up to 8 GPUs to any VM to create your own GPU farm. Connect your VMs to a variety of storage solutions for further cloud processing, or synchronize data with your local file system. GCP’s software-defined network means you won’t create bottlenecks by throwing too much on the farm at any given time. Speeding up render time like this has already been useful for GCP media and entertainment customers to scale faster with less wasted resources. And it’s all done securely, as GCP undergoes regular audits by the MPAA and Independent Security Evaluators.
Today, we are happy to announce our new guide, Building a hybrid render farm on Google Cloud Platform. Take a look to learn how to securely extend your on-premises render farm to Google Cloud. It covers connectivity options, how to synchronize your data, software licensing, and best practices to deploy and monitor resources. The guide also covers industry-specific topics such as production and asset management, archiving, and render queue management integration.