Though initially founded as a color finishing house, Nice Shoes has steadily expanded its scope of capabilities to include more CG-intensive projects, including visual effects and immersive experiences. As the studio has taken on more complex work, it’s evolved its pipeline to keep up, adopting Deadline compute management software from AWS Thinkbox in early 2019. Shortly thereafter, Nice Shoes began determining the best approach to refresh its on-premises render farm and found a solution in the cloud on AWS.
“Deadline is integrated with AWS almost out-of-the-box so that made transitioning to the cloud that much simpler,” noted Nice Shoes Senior Systems Engineer Phil Wortas. “I worked with the AWS Thinkbox team on our proof of concept in May and the whole process was seamless. They helped me get rolling, and the entire experience was painless. The timing also ended up being fortunate because it turned out we needed to jump right into production on a project with a tight turnaround.”
As Wortas was preparing to leave Nice Shoes’ New York office on a Friday evening this past summer, he received a call from one of the studio’s executive producers who was concerned that the team wouldn’t have enough render resources to make a Wednesday morning deadline.
“The client needed about 40,000 frames rendered in less than three days, and locally we only had eight nodes, plus the artists’ workstations. Do the math, and that’s not nearly enough compute power, so we went into panic mode,” Wortas explained. “At first we talked about getting more nodes on site, but that introduces a host of scaling hardware challenges, like rack space and cooling power. The freight elevator was also closed for the weekend, which meant we couldn’t even get equipment into the building if that was the route we wanted to go. We’d been testing cloud rendering on AWS with great results so I suggested we try using it for this project.”
Jumping from proof of concept into production required slight workflow modifications, but the Nice Shoes team was able to get onboarded and working nearly instantly. First, Wortas created an Amazon Machine Image (AMI) to spin up the virtual machine (VM) render node. “Amazon’s infrastructure gives you ownership over your instances when you spin them up, which is extremely helpful. It means you can go into the AWS Portal via Deadline, grab the keys, and connect to the nodes through SSH (secure shell), then do whatever is needed.”
From the AWS Portal in Deadline, Wortas spun up instances, configured them as needed, saved, and then replicated them. Next, he purchased credits for on-demand Nuke and Deadline licenses from the AWS Thinkbox Marketplace, which gave him on-the-spot access to the software per-minute, and saved Wortas from licensing considerations. He configured two previously siloed servers, then applied maintenance scripts, and by Friday night, the team was running test frames on AWS. Wortas also set up bring-your-own (BYO) licensing in Deadline so that the artists could render with Nice Shoes’ 40 in-house permanent seats before drawing from usage-based licensing (UBL) credits.
Frames began pouring in on Sunday, and Amazon Elastic Compute Cloud (EC2) Spot Instances, using the c5.18xlarge instances featuring 72 cores and 144GB RAM, powered the rendering workloads. “Spot Instances are dramatically cheaper since they’re essentially extra space on the server. The c5.18xlarge instances in particular seemed to provide the right balance between cost and performance,” Wortas noted.
Ultimately, the project delivered on time and drew internal attention at Nice Shoes. Wortas shared, “After other artists saw what we accomplished with AWS in only a weekend, they were all clamoring to use it. Naturally, we’ve since brought AWS more broadly into our pipeline, where it’s served a range of projects, including one recently rendered in Arnold and with Houdini fluid simulations. The fact that we were able to pivot almost immediately is a testament to flexibility of AWS.”
Moving forward, Wortas plans to apply lessons learned from Nice Shoes’ first cloud-powered projects and use rendering with AWS to further optimize the studio’s workflow through automation tools in Deadline, like the Spot event plug-in. He’s also exploring a full Studio in the Cloud workflow, which makes use of virtual workstations, rendering, and storage on AWS.
“Workstations are a great place to use cloud scalability, and there are pretty powerful instances available on AWS already. The industry is trending away from on-prem, and we’re already exploring going further with the cloud. Our half rack may be fine for now, but we’re in a heritage building and as we grow, it will be difficult to get additional cooling, power, and machines to support more artists. With AWS, there’s never a concern about scaling.”