Dreamworks' storage tiers help animate movie making

Dreamworks uses an eclectic blend of storage tiers and technologies to handle billions of files and petabytes of data required to make its films.

For Dreamworks Animation SKG, making movies doesn't involve going to remote locations or shooting scenes in a Hollywood studio. The action in its movies takes place inside computers where artists create the animated characters and scenes, which requires storage tiers that can handle stringent performance requirements, as well as big data.

The recently released animated movie, The Croods, was merely one of many films the Dreamworks' teams of engineers have been working on. Dreamworks typically releases two or three films a year, and each one takes from three to five years to complete. That means there are about 10 movies in production at any time. Each movie can have approximately 500,000 files, so that comes to about 5 billion digital files in use.

The animation studio uses four data centers -- two in California, one in Las Vegas and another in Bangalore, India. The render farm has 20,000 stations. Dreamworks has more than 5 PB of disk storage, and its archival tier swells by around 200 TB to 300 TB annually, according to Derek Chan, head of Dreamworks' Technology Global Operations.

"We produce billions of files," Chan said. "Our end output is data that comes out in [the form] of images. We see the amount of data we're storing increasing dramatically every year."

Dreamworks has an eclectic mix of storage, with traditional SAN and network-attached storage (NAS) products, as well as server-side flash caching and storage in the cloud. Dreamworks has a long-standing relationship with Hewlett-Packard (HP), and runs most of HP's main storage platforms -- StoreServ 3PAR T400 and V400, HP P200 and EVA 6400 SANs, StoreAll 9730 for archiving and NAS, and StoreOnce B6200 for deduplication and data protection.

But Dreamworks' data centers also include NetApp FAS6280, 6240, 6210, 6080, 3170 unified storage, Avere FXT 3500 NAS acceleration systems, a Panzura Cloud Storage Controller that moves data to the HP storage cloud and Fusion-io PCI Express flash cards inside servers for a performance boost. There is also tape for archiving.

"We're not beholden to any vendor," Chan said. "We're looking for best-in-class solutions."

Chan said Dreamworks divides its data into three tiers of storage: high performance, near-line and asset protection. The high-performance tier includes data in production and business applications such as its Oracle database, near-line is for bulk storage, and asset protection is for archiving. The high-performance tier uses 3PAR and NetApp arrays with Avere for caching. Near-line storage includes the StoreAll and StoreOnce along with HP and NetApp storage. Older disk systems, the cloud and tape are used for asset protection.

Dreamworks' high-performance storage needs to keep up with thousands of artists working on files.

"All of our artists are simultaneously hitting the same infrastructure," said Kate Swanborg, technology executive for Dreamworks. "If they're waiting for the technology, that's not good for us."

The asset-protection tier stores Dreamworks' completed movies that may have to be brought back for sequels or to be upgraded when new technology is developed, such as when the Shrek movies were re-released after stereoscopic 3-D came out. "We'd rather not go into tape because the time to restore the data off tape is prohibitive," Chan said. "We'd rather have it accessible online."

Dreamworks has about 50 TB in the HP cloud, using Panzura's appliance as an NFS mount point. "We keep multiple copies of data for our archive," Chan said. "One is on tape, one on spinning disk and one that we put out in the cloud. It doesn't matter how it's stored, as long as it's secure and we have access to it."

Chan and Swanborg said their storage systems typically have a three- to four-year life span. With so many systems in production, their team spends a lot of time evaluating all the options on the market.

"We're stringent about it," Swanborg said. "While the world sees our product as a movie, our product is data. We look to our engineers to be highly diligent to evaluation [on] whatever opportunities are there."

Chan said interoperability between storage systems isn't much of a problem because Dreamworks' artists work primarily with NFS files. "We try to stay away from proprietary and stick to open source or open standards so we can go to multiple vendors for support," he said.

With its rate of data growth, Dreamworks is looking to expand its deduplication and compression technologies. Chan said his team uses NetApp's built-in dedupe and is considering HP's StoreOnce for primary data.

"It's interesting if it's built right into the storage devices so we don't have to think about it," he said of dedupe.

Dig deeper on Tiered storage

Pro+

Features

Enjoy the benefits of Pro+ membership, learn more and join.

0 comments

Oldest 

Forgot Password?

No problem! Submit your e-mail address below. We'll send you an email containing your password.

Your password has been sent to:

-ADS BY GOOGLE

SearchSolidStateStorage

SearchVirtualStorage

SearchCloudStorage

SearchDisasterRecovery

SearchDataBackup

Close