AWS enables Outpost VFX to service new clients anywhere in the world in hours
When he founded Outpost VFX in 2012, CEO Duncan McWilliam never thought that his studio’s on-premises computing infrastructure would max out the building’s electrical system—or that he would be asking the landlord for more power.
Outpost VFX creates visual effects for films and episodic television. Some of its recent credits include the Amazon Prime series, “The Lord of the Rings: The Rings of Power” and “Wheel of Time;” the Skydance series, “Foundation;” “Fast X” for Universal Pictures; and the Netflix film, “Slumberland.”
Besides using significant power as the business grew globally, the system reached its productivity limits. “We had expanded to Montreal, Los Angeles, and Mumbai, and we realized that we needed something more flexible than on-prem compute,” McWilliam said. “The question was, how do we scale the business without paying significant capital expenditure that we didn’t have pre-investment?”
The answer was the cloud. And not just for some tasks—for everything—with AWS. CTO Tim Chauncey explained, “We had some cloud presence previously that helped us to support our different sites and deliver on our projects. But we made a conscious decision to go fully into the cloud. We saw that as the path forward, so we made a decision to go all in.”
That’s how it started. Engaging with AWS connected Outpost with Konsistent Consulting of the Amazon Partner Network (APN). Together, they did intense operational planning and equally comprehensive financial modeling. With expert guidance from AWS and Konsistent, Outpost moved close to a thousand users at four studios and a massive amount of data to the cloud in just two and a half months—all while in full production and delivering to clients with minimal interruption.
Connecting global teams in the cloud
Today, Outpost has five studios, having added a London location, all functioning as a single studio in the cloud, with all workstations and storage on AWS. “We built the entire operation to feel like one global studio,” said McWilliam. “The more we can make everyone feel connected, the better.”
Moving to a full studio-in-the-cloud approach, McWilliam said, “has allowed us to be more flexible and more iterative with our clients. We’re able to say yes to them a lot more.”
Chauncey elaborated, “We can do larger simulations, more complex sequences, bigger projects, more renders, and more shots. We can scale up as needed. Previously, it would’ve been a case of, ‘well, we could do this sequence but we wouldn’t be able to do that one.’ Or, ‘we could do these shots, but don’t have the capacity to take on these others.’ Now we can scale relatively simply to accommodate anything we want to do. When there are shots clients might be struggling with, we can take them on.”
Shifting to an operational expenditure mindset
One of Outpost’s core values is that they operate like one business regardless of location. McWilliam’s vision was for a seamless, single network, where every user has the same experience globally. Key to achieving that was consistency. All studios use the same cloud-based hardware: either G4dn or G5 workstations, each running NVIDIA GPU EC2 instances for Houdini, Nuke, Maya and other standard packages artists need. Then, if an artist requires more compute power or RAM, they can scale that up within minutes via AWS.
“From an artist’s point of view, the fact that they don’t have a machine under the desk is the only thing that really tells them that they’re in AWS,” said Chauncey.
This flexibility is new to artists, and there is a certain temptation to throw every project to the cloud with the highest settings, spinning up thousands of machines. Outpost has put mechanisms in place to address that. “We’ve built a lot of tools around our core infrastructure to monitor what’s happening and to see where resources need to be allocated.” They also have a dedicated resource department to make sure that resources are balanced; that they’re not overusing in some places and under using in others. “We’re moving into much more of an OpEx world where all of these resources need to be managed and monitored correctly. It’s that mindset, that shift in ideology, that needs to be adjusted,” Chauncey said.
Rendering and storage elasticity
In terms of the render environment, the studios run thousands of instances scaling from very small instances for ad hoc tasks to 96 core CPU instances for tasks like FX renders and explosions, water simulation and other high performing workloads. “That can be scaled dynamically, and we have the capability to reach high numbers of render instances,” Chauncey said. “But the fact that we can pull the farm right back down is the important part. So, as those renders burst, they can scale out, and artists can quickly iterate their work, but the business gets the benefit of having that reduced scale when it’s not in use.”
With so many artists worldwide hitting storage with high render compute requests, Outpost needed a mechanism to handle that load and maintain performance. They partnered with WEKA IO for their storage solution. “The great thing about it is, it scales really well,” Chauncey said. “On the other hand, it’s so easy for artists just to keep adding more and more data.” This process too is subject to rules set for the OpEx model that make sure the studio isn’t creating redundant data.
Currently, Outpost uses four instances of AWS Thinkbox Deadline render queue management software for additional orchestration. By automatically invoking compute resources based on the work in the render queue, and then shrinking them again based on workload demands, this simplifies the management of cloud resources.
As Outpost VFX began scaling to a global, multi-office presence, the company harnessed computing resources across multiple regions, and looked to Deadline to streamline the process further with enhanced render capabilities for artists based around the world. “Deadline has greatly streamlined the management of our cloud render resources,” explained Chauncey, “and with the newly developed multi-region support in the Spot Event plug-in, we are actively exploring ways to further enhance scalability and optimize efficiency.”
Managing change and costs in the cloud
McWilliam offered some of his team’s key learnings from orchestrating the massive cloud migration for other studios that are considering moving more of their production from on-prem to a cloud-first workflow. “Planning is intense. Understanding costs is very complex. Take it slowly.”
Chauncey added, “Keep a broad view of what you’re trying to achieve; don’t get caught up trying to work out all the tiny nuts and bolts.” Both pointed to the need to think—and operate—differently once studios have migrated to a cloud infrastructure.
McWilliam advised, “Put a shutdown script in place for when people leave their boxes on when they go home at night. Also, make sure your disciplines are working symbiotically worldwide and in a structured manner. The point of the cloud is you need to manage it and treat it with real respect, then you start to see the cost benefit. Be careful it doesn’t go the other way.”
Chauncey noted that because of the consistency in how Outpost has structured its studios, “We’re at a place now where we should technically be able to spin up a brand-new site in a new region within a very short time frame. If we wanted to, we could service a client in a completely different part of the world, spin up a site and we could have that up and running in a matter of days.”
McWilliam and Chauncey spoke to the critical importance of support, particularly from AWS, in transforming their business with the cloud. “AWS has vast experience in migrating not just visual effects companies, but any company into cloud architecture. The support, the talent AWS put us in touch with, the resources made it a sensible choice.”