361
Shares
87
0
274
0

Uniting the efforts of creative workers all over the globe

In this article...

  • DreamWorks future-proofed its enterprise when it invested in multi-core processing technology years in advance of its competitors
  • Due to an automated Cloud infrastructure, technical errors can be flagged before impacting artists' workflows

Derek Chan is head of global technology operations for DreamWorks Animation where he is responsible for providing the strategic vision, planning and implementation of the studio’s computing infrastructure.

DreamWorks Animation, a global leader in the creation of groundbreaking computer-generated (CG) animation, employs hundreds of artists across three studio locations. For DreamWorks, uniting the efforts of creative workers all over the globe requires an IT infrastructure that can withstand extreme demands for speed, power, flexibility and automation.

How are your computing demands different than a typical enterprise?

It may not seem obvious that an animation company requires a high performance compute environment. However, each one of our films is made up of half a billion digital files, requires hundreds of high-end workstations, storage for hundreds of terabytes of data and the orchestration of tens of millions of compute hours. With seven to 10 films in production at any given time, the scale of our business is significant and our processing power demands — from the data center to the desktop — are extraordinary.

“Adding to the weight of our data, our creative environment requires cross-site collaboration.”

On the desktop, our artists are driving highly specialized programs that are incredibly compute-intensive. If we provide an animator with a workstation equipped with 16 cores, the animator will use all 16 cores available to him. Conversely, the typical office worker might only use spreadsheets, Microsoft Word and email.

The differentiator for us lies in the processing power required to support our infrastructure. DreamWorks Animation is a high compute environment, hosting more than 15,000 cores across three data centers to support studio locations in Glendale, California; Bangalore, India; and Shanghai, China.

Adding to the weight of our data, our creative environment requires cross-site collaboration. Our artists are constantly communicating and sharing work with their colleagues halfway around the world. The need for instant and seamless collaboration requires us to have an infrastructure that is agile, flexible and innovative.

What’s at risk if the infrastructure lags behind the artists?

Simply said, we want our artists to focus on their art, not on the technology that drives it. The creative process works best when an artist is able to stay “in the flow” without disruption. When technology stops working, slows down or requires a fix, the artists are suspended from their train of thought, which inhibits the creative process. If our artists are able to focus on their craft rather than worrying about the technology driving the tools, the result will be higher quality work.

 

How does supporting this flow state translate into a technical requirement?

Our big focus when we build tools for the making of CG animated films is: How do we reduce the wait time for our artists?

How does pre-computing work in animation?

Six years ago, we learned that the technology industry would experience a shift from single-core to multi-core computing. We quickly realized how powerful that shift could be to our processing power, so our engineers set out on a massive undertaking to re-architect our proprietary software applications to optimize for multi-core computing.

“It’s really about having a data center, or a Cloud environment, that is self-correcting in a way.”

With pre-computing, for example, an animator might be working in one thread looking at one application. Meanwhile, that application is working in the background to predict what the animator might do next. Then the software will start to pre-compute what would be needed if the animator completed certain operations.

For the end user, it is very much like typing into a smartphone. As you type, the phone is predicting what you might type next. Pre-computing in animation is the same concept on a much larger scale, since generating the next image is significantly more complex than predicting the next word.

The results have been unbelievable. Our artists are now working in ways never thought possible. Now that our software is enabled to leverage multi-threaded computations, the artists’ workflows have completely transformed to be intuitive, iterative and in real time.

How does automating infrastructure help you accelerate your move toward pre-computing?

It’s really about having a data center or a Cloud environment that is self-correcting, in a way. An automated infrastructure is out there looking for mistakes and trying to flag errors before they impact the artists. For example, if a process needs to go through 100 steps but slowed down on step 55, an automated infrastructure will give us visibility into step 55. What’s happening? Are we seeing a bottleneck there?

The bottom line is that the better enabled we are to manage our infrastructure and deliver applications, the better we can deliver services to our artists. With three studio locations worldwide, agility and flexibility are essential to promoting more collaborative workflows across multiple sites.

How Much Have You Read?
0%
Complete
Subscribe to the HPE Matter NewsletterSubscribe to the HPE Matter Newsletter
Listening