The new 3D animated movie "The Croods" may be about a stone-age family, but DreamWorks said it is by far its most sophisticated production to date, topping all others in compute cycle hours.
The movie, out in theaters last Friday, required a whopping 80 million compute hours to render, 15 million more hours than DreamWorks' last record holder, "The Rise of the Guardians."
Between 300 and 400 animators worked on "The Croods" over the past three years. The images, from raw sketches to stereoscopic high definition shots, required about 250 terabytes of data storage capacity to make, according to Kate Swanborg, head of enterprise marketing at DreamWorks.
"Storage sounds like a passive word, but it's an active part of our infrastructure. Artists are actively accessing around the globe in our three studios various parts of data every single minute of the day," Swanborg said
After completing a film like "The Croods," Dreamworks archives about 70TB worth of data, images that can be reused over and over again in future productions—things like background art or plants.
So much data
"One of the things not always evident is how much data we're dealing with in the creation of one of these films," Swanborg said. "We release two to three films a year, and at any given time we have 10 films in production, and they're in all different states of production. And each film will have generated half a billion files."
Each animated character in a 3D movie has up to 2,000 control points or features that can be manipulated by an animator. And, each character takes six months to craft, said Swanborg. A competed film has more than 250 billion pixels in it.
When the movie industry moved from producing 2D to 3D high-definition movies over the past decade, the data required to product the films increased tremendously. For DreamWorks, the amount of data needed to create a stereoscopic film leaped by 30%.
DreamWorks has standardized much of its IT infrastructure on Hewlett-Packard's HP BladeSystem c-Class server blades—3,000 of them—which consist of preconfigured compute, storage and network architecture. It also uses HP NAS and HP's 3Par storage arrays. The only exception to the HP hardware is a small amount of storage from Hitachi Data Systems and NetApp NAS.
DreamWorks' five petabytes of disk storage is tiered, from solid-state drives and volatile cache for high performance applications, to 15,000rpm hard drives and high-capacity SATA drives for nearline and archival storage.
Networking around the globe
DreamWorks employs animators around the world to work on its films, which means it has to supply enough network bandwidth so that at any point in time and artist can collaborate across the globe on any ongoing project.
The network connection between its Glendale and Redwood City studios consist of two 10Gbps ethernet networks with automatic fail over, said Dave Thomas, storage and backup supervisor for DreamWorks Animation.
Those networks are tied into a central hub that offers up to 500MB/sec to artists in Bangalore, India, where, in 2008, DreamWorks established a special unit within Technicolor named DreamWorks Dedicated Unit.
DreamWorks has a "render farm" of servers made up of about 20,000 processors. The image rendering jobs are broken up into small pieces, distributed out to the server farm, and are later recompiled to create the final images for a film.
"When we look at a film like 'The Croods,' we don't think it's better or worse if it takes more render hours or less," Swanborg said. "Obviously, how it's beloved by audiences and how well it does at the box office is our measure of how great that product is.
"Our philosophy is to create an infrastructure and engineering approach that allows us to say yes to our filmmakers' visions," she added. "If they have a vision that requires a certain amount of compute, we don't want technology to get in the way of that, we want the technology to enable that."
This story, "The computing muscle that DreamWorks needs to create 'The Croods'" was originally published by Computerworld.