Method Studios Takes Cloud Rendering to the Next Level with AWS
Scales Local Capacity 4X Using EC2 Spot Instances for “Jumanji: The Next Level” VFX
The artistry of global visual effects company Method Studios can be viewed regularly on screens of all sizes, whether in blockbuster films and episodic television series or high-profile commercials, title design, and beyond. With locations around the world, including Los Angeles, Melbourne, Montreal, New York, and Vancouver, the company has a suite of tools to allow for fluid collaboration between facilities, though each site also customizes its workflow based on project needs. Since becoming cloud-enabled in 2017, Method Melbourne frequently taps Amazon Web Services (AWS) to scale its render resources, including for Sony Pictures’ “Jumanji: The Next Level.” The studio was enlisted to transform bluescreen sets into complex digital environments, then seamlessly integrate actors for a tension-filled climax featuring a snowy fortress, a massive blimp, and a flying horse. Led by Method VFX Supervisor Glenn Melenhorst, the project required the creation of 280 VFX shots, 164 of which were completed using AWS.
“We’ve used AWS on almost every project we’ve worked on in the last two years and it’s been vital for the final delivery push,” noted Method Melbourne Head of Information Technology Adam Jones. “Our on-site render farm is still fairly robust, but we are no longer investing in local resources and find ourselves going to the cloud more and more since it is available to us on an as-needed basis.”
When offloading to the cloud, Method Melbourne almost exclusively uses fleets of Amazon Elastic Compute Cloud (EC2) Spot Instances that run through the Sydney AWS region. Working on “Jumanji: The Next Level,” Method ran 147,904 tasks on AWS over four and half weeks using a mix of M5, R5, and C5 instance types. Hitting 480 instances at peak production, the facility was able to scale its internal, 10,000-core farm 4X with AWS.
“The way we look at rendering and hardware completely changes when using the cloud,” Jones explained. “When you render in-house, you spec out what you think you need for the next three or four years in terms of best hardware for price and performance. With cloud rendering, you can customize machine specs by render task, and also use resources concurrently, which is more efficient for licensing. It completely shifts your mindset when there is no price difference in spinning up one machine for 100 hours or 100 machines for one hour. Quicker render returns help with iteration and the creative process, and give clients more freedom for artistic decisions.”
Method Melbourne began using AWS following its native integration with Deadline render management software, developed by AWS Thinkbox. “We have been using Deadline for 15 years and have a longstanding relationship with the Thinkbox team. They’re always very gracious with their time and were integral in getting us up and running in terms of planning our networking and infrastructure, and setting up the Amazon virtual private cloud. With the AWS team working along-side our Head of Systems Jon Stanley and Head of R&D Grant Adam, we had a production-grade proof-of-concept running within a month,” shared Jones.
He continued, “From a security standpoint, AWS ticks all the boxes: they’re MPAA compliant, and we’re able to securely build what we need in terms of pipeline. Additionally, regarding cost, it’s an elegant financial solution. It’s much easier to work with operational expenditure on render than to purchase new hardware, as you only use it when you need it, which is fantastic from a production and financial reporting point of view.”
Based on its success using AWS for rendering, Method Melbourne is now exploring how virtual workstations might benefit its workflow. Jones concluded, “Virtual workstations are our next step. We’re looking at spinning up machines in Windows. We mostly use Linux-based workstations but there are times when our artists need to use applications native to Windows. Rather than having to use two local machines, the ability to spin up a virtual workstation in whatever specs are needed is quite attractive.”