Bardel optimised its animation pipeline for cloud implementation, hiring the cloud consultants at DeadDrop Labs to help integrate Thinkbox Deadline and launch the system on AWS.
Bardel Entertainment in Vancouver has been working as a 2D, 3D and hybrid animation studio since 1987. Working alongside Netflix, Disney+, HBO, Nickelodeon, Cartoon Network, DreamWorks and Warner Brothers, their team creates original primetime programming for all ages as well as content for streaming platforms.
To start preparing its operations for the future, Bardel recently overhauled and optimised its pipeline for cloud implementation, hiring the cloud consultants at DeadDrop Labs (DDL) to help launch the system on Amazon Web Services (AWS).
Led by Bardel’s Vice President of Technology Arash Roudafshan, the studio’s pipeline upgrade also serves as a foundation for its expansion into more feature film projects, and to enhance multi-site collaboration. Considering Bardel’s fast-paced episodic production schedule and small engineering team, Arash asked DDL to design and manage the transition. Meanwhile, the Bardel teams collaborated and were able to test and approve the new pipeline in less than a month. The new, hybrid system was fully operational in four months.
Designing the Transition
Working with DeadDrop Labs helped Bardell negotiate the public cloud environment faster and more efficiently than they could have on their own, ensuring security with TPN (Trusted Partner Network) approvals.
“Using Terraform, we built a new account and organisational structure from day one, rather than having to undo decades of legacy hierarchy first. We were able to go to market much faster, and since the strructure is built with an automated backend on AWS, we can now scale as our needs change.”
Terraform – developed by AWS partner HashiCorp – is an infrastructure-as-code tool used to define cloud and on-prem resources in human-readable configuration files that clients can version, reuse and share. Taking this approach makes it possible to use a consistent workflow to provision and manage an infrastructure throughout its lifecycle. Terraform can manage core components like compute, storage and networking resources, up to higher-level components like DNS entries and SaaS features.
Making Connections
DeadDrop Labs helped Bardel implement its own multi-account AWS environment based on workstations and virtual desktop infrastructure (VDI), rendering, networking and shared services. On-premises, the networking and connectivity have been centralised using a 10GB AWS Direct Connect network connection with multiple private virtual interfaces (VIFs) for redundancy,
This network is connected to a Direct Connect gateway, which connects the virtual private clouds (VPCs) across Bardel’s AWS accounts and sets up peering to allow devices on one network to exchange traffic directly with devices on another network.
“We want to broaden our company horizons and capacity to take on more, larger projects, but deliver at a consistent high quality. The pipeline we’re developing will allow us to scale with global AWS resources and, with that support, enter any market that makes sense,” Arash said.
“From the artists’ standpoint, they don’t need to concern themselves about where the resources are, just that the machines work and are optimised for their use case. Any areas where we can improve the artist experience – like reducing load times – is an advantage, on the assumption that they’ll be able to work more fluidly and turn around better work, faster.”
Render Management
AWS Thinkbox Deadline scalable compute management software, currently running on-premises, is used to manage rendering. As Bardel submits render jobs, they are queued locally and on artists’ workstations, with data stored on-premises and, as described, accessed over Direct Connect. When Bardel’s compute needs outstrip this system’s capacity, it scales to AWS using Deadline’s built-in Spot Event plugin.
AWS makes temporary low-cost Spot compute capacity available on the fly. Deadline is configured with various groups and pools to take advantage of different Spot fleet sizes and instance types across multiple availability zones in the selected region, meeting capacity needs within budget.
Arash said, “Because Bardel was established some time ago, we have a lot of core legacy backend hardware that isn’t ideal for modern workflows. Upgrading that infrastructure has been a priority so that we’ll be ready to capitalise on new developments, including Deadline features like log collection and metric management, when scaling up and down.
“By taking better, more responsive advantage of Deadline reporting, automated services can be built and share production data with supervisors and managers so that they can make decisions and forecast based on actual information, upfront. If you only rely on brute force, everything suffers, and you run into budgetary issues.”
Workflow and Artist Support
Bardel also takes a hybrid approach to workstations and VDI, using the cloud or machine rentals for extra capacity. The studio’s primary creative applications include Autodesk Maya with Foundry Katana and NUKE for 3D animation, and Toon Boom for 2D animation. Autodesk ShotGrid is used for production tracking. Beyond these products, Bardel uses a variety of proprietary tools, as well as some real-time game engine software.
Bardel’s Vice President of Technology Arash Roudafshan
Whether on- or off-site, artists typically log into secure machines remotely for production. Since Bardel has adopted a new model of purchasing bigger servers with graphics acceleration, they can divide the capabilities virtually so that multiple artists work on the same machine, scaling up the studio’s capacity while in effect reducing the physical footprint.
Dual VDI Environment
The VDI environment is comprised of two separates AWS accounts, one for development and another for production. This arrangement allows the software, processes and connectivity to be fully developed and staged with zero impact, crosstalk or access to the production environment. Once properly tested and validated, the machine images for that set-up are pushed to the production
Along with running creative applications in production, about half of each machine’s compute is dedicated for daily renders. Bardel’s render and VDI environments are connected via VPC peering, linking a shared services VPC used for tooling and common utilities, such as managing license servers, and a storage VPC containing cloud storage resources shared across multiple accounts.
"Working with DeadDrop Labs team when we created the foundational layers was a huge time savings for our staff,” said Arash. “We were able to quickly build an AWS-enabled pipeline environment to meet our unique specifications that scales really well – I could launch a new region across the world and have artists connected in 20 minutes.
I look forward to our ongoing development efforts and how our pipeline will evolve with technology and industry shifts. Through all of our growth, our commitment to delivering high quality content remains the guiding principle.” aws.amazon.com
