Written by Geoff Newell
How an International VFX House Leveraged HPC to Meet Time-Critical Deadlines
Appsbroker’s client is an international media and entertainment organisation, providing products and services to enable the production of animated motion pictures that are loved across the world.
As innovations such as photo-realistic 3D animation and 4K content continue to drive up the quality of high-definition rendering, the creative and technical resource demands have never been greater. Technological advances have also significantly reduced production times and enabled more agile working, with some studios and filmmakers now expecting to render effects and animations on the same day as filming.
To meet the large volumes of demand from its customers, the solution required a large support team of animation wranglers and engineers, globally distributed across different time zones. It was extremely difficult to forecast usage due to the high-paced nature of the industry, meaning VM capacity could quickly be reached and exponentially increase the waiting times for short-notice, resource-intensive render jobs.
This stretched the availability of human and technical resources as the teams tried to manually balance budget and speed when queuing jobs, while also being constrained by a finite number of available VMs. Meanwhile, end-customer requests continued to increase for rapid rendering done within exacting timescales and stringent budgets, forcing the client to find a way to adhere to modern best practice and remain competitive.
The client looked to Appsbroker and Google Cloud for their joint experience in successful delivering cloud-based high-performance computing (HPC) projects.
Appsbroker’s engineering team developed a bespoke solution to integrate with the client’s VFX scheduler. This connected the existing premises-based solution to Google Cloud Platform (GCP) and the Intel Xeon Scalable Processor (Skylake) powered Google Compute Engine (GCE) VM resources required for performant rendering. Google Cloud’s fleet of Skylake-powered VMs allowed the client massive compute on demand to ensure exacting movie deadlines could be hit on time and on budget.
A flexible solution that utilises existing on-premise servers and that can quickly benefit from the flexible nature of Google Cloud compute optimises the client’s existing investment in hardware. The ECCoE team at Appsbroker specialise in helping clients adopt dynamic compute for the largest workloads operating in a hybrid fashion.
The integration with the scheduler automatically takes on additional render jobs through GCP on a pre-configurable basis, with parameters controlled by the client’s team. These can include rate of VM provisioning per hour and enable oversight and control of the length that certain jobs will take to be rendered against the additional usage cost. VMs are provisioned based on the criteria and then automatically removed once inactive to prevent excess use of capacity and remain cost-effective.
In this way, not only can the solution scale up on-demand to process last-minute high-resource render requests without going over budget, but it also makes general over-capacity rendering predictable and easily manageable – without placing additional demand on the technical support team. This means that end-customers are able to consistently enjoy rendering of the quality they need in the timescales they want, better navigating the challenges of delivering productions within tight budgets.
The development of the render infrastructure provided the organisation with far greater control over project delivery. As part of a true hybrid cloud infrastructure, the client benefits from the scalability of cloud without impacting on the CapEx investment into its on-premises data centre. Using Intel Skylake core VMs for all cloud rendering projects has seen compute power increase by over 30%, at a lower cost than alternative CPU architectures.
Over the course of a single weekend, the solution was able to provide an additional one million cores of capacity through GCP, enabling rendering of unprecedented scale for a time-critical production. The ability to deploy HPC workloads at mass scale across its on-premises and Google Cloud Platform instances has significantly improved the organisation’s ability to respond to the myriad challenges posed by the pace and demands of today’s film industry.