IBM > Case Studies > HPC chops rendering time for world-class animation studios

HPC chops rendering time for world-class animation studios

Customer Company Size

Large Corporate

Region

America

Country

United States

Product

IBM® Platform™ LSF®

Tech Stack

HPC (High Performance Computing)
Linux

Implementation Scale

Enterprise-wide Deployment

Impact Metrics

Productivity Improvements
Brand Awareness

Technology Category

Infrastructure as a Service (IaaS) - Cloud Computing

Applicable Functions

Discrete Manufacturing

Use Cases

Mass Customization
Digital Twin

Services

System Integration

About The Customer

The customer is a major Hollywood studio that produces high-quality family entertainment through the use of computer-generated (CG) animation. The company's world-class creative talent and technological capabilities allow it to deliver great stories, breathtaking visual imagery, and a sensibility that appeals to both children and adults. The studio is known for its high production values and the complexity of its animated films, which require significant computing power to render. The studio operates two facilities, each with approximately 2,000 CPUs, and employs up to 300 people, including animators and lighting technicians, who submit work and run models against the compute farm on a typical day.

The Challenge

The Hollywood studio produces high-quality family entertainment through the use of computer-generated (CG) animation. The company’s world-class creative talent and technological capabilities, allow it to deliver great stories, breathtaking visual imagery and a sensibility that appeals to both children and adults. However, the amount of computer time required to render animated films grows with each technological improvement. As such, HPC in the form of massive computer grids is a critical technology in CG film-making. The studio’s latest blockbuster, a children’s film required the equivalent of over 17 million processor-hours of computing time on some of the world’s fastest commercially available processors.

The Solution

The studio chose Platform LSF as the heart of the HPC environment that underpins the render-farm grid. Platform LSF manages 4,000 CPUs in two facilities, and vast numbers of jobs in a virtual computing environment that never goes down. Platform LSF allows all of the processors to act as a single, large shared render farm operated as two large Platform LSF clusters, each composed of approximately 2,000 CPUs. In addition to the dedicated server farm, animators and designers also have several hundred desktop computers running Linux at each major facility. Up to 300 people, including animators and lighting technicians, submit work and run models against the compute farm on a typical day. Overnight the desktop systems join the render farm, contributing several hundred more computers to the powerful grid computing infrastructure.

Operational Impact

The studio was able to complete the complex rendering job in time to get the film released before the summer blockbuster season.
The studio has developed its own “currency” for the use of the grid called “RU”s or “render units.” Based on internal capacity planning activities, a number of render units are distributed between each competing feature and work with feature directors to allocate rendering budgets within each film.
Using this approach, Platform LSF ensures that each production is getting their prescribed “fair share.” This allows the studio to control how resources are shared by different production efforts and different functions competing for resources on the grid according to predetermined allocations.

Quantitative Benefit

Overnight the render farm has an average utilization of over 97 percent, including the desktop systems pressed into service by the Platform LSF cluster.
On average, between 20,000 and 30,000 jobs run overnight, but as many as 100,000 jobs can run during the busy “crunch” time that precedes a major feature release.