Important Looking Pirates Case Study

Important Looking Pirates (ILP) is a visual effects (VFX) studio based in Stockholm, Sweden. Its founders Niklas Jacobson and Yafei Wu wanted to bring the same “passion, perfection, and development” to their local scene as they had found working in VFX houses in the U.K. and U.S. Since its foundation in 2007, the company has grown to include nearly 50 employees, around three quarters of whom are artists creating content for TV, film, commercials, virtual reality, web, and games for an international client base including some of the industry’s biggest names.

One of the greatest computing challenges in VFX is the rendering of images that comprise the final product. To do this, ILP has an in-house render farm of around 100 nodes. However, like any agency, it often has large variations in the size of workloads. Cofounder Niklas Jacobson says, “Thanks to the success of our work in the broadcast market, we were getting a lot more contracts. Things become especially problematic when two projects are wrapping up at the same time. In this situation, we can take on freelance artists to deal with spikes in demand fairly easily, but scaling our infrastructure with the same flexibility is an issue. Obviously, we’d prefer not to invest in permanent hardware for short-term needs because achieving maximum utilization across long periods is challenging.”

It made sense for ILP to look to the cloud for a solution. This way, the company could pay for compute resources on demand, and stop paying once jobs were completed. It had been using Amazon Web Services (AWS) to render on a small scale, but the solution wasn’t workable with large scenes, as Jacobson explains: “Rendering scenes today involves transferring huge amounts of data. To give you an idea, a single frame sent to the render farm can easily hold 2 gigabytes of data. And during a busy month, we may push more than 20,000 render and simulation jobs through our systems. We had used cloud services, but the transfer speeds were a bottleneck, so we couldn’t rely on the system during busy times.”

ILP discovered that Avere, a company on the AWS Marketplace, offered a solution to its problem. “When Avere came up with its vFXT Edge filers for Amazon EC2, we saw possibilities to make bigger rendering jobs in the cloud at a speed that was just not viable before,” says Jacobson.

Render jobs are managed internally, and artists can send work to the cloud or local nodes using exactly the same process. ILP connects to its AWS cloud environment via the Amazon Virtual Private Cloud (VPC). Avere’s vFXT Edge filers run on Amazon Elastic Compute Cloud (Amazon EC2) instances alongside the studio’s rendering applications, which use Amazon EC2 Spot Instances, offering up to a 90 percent reduction on typical on-demand prices. The Avere cluster automatically moves scene data to the cloud, providing extremely low-latency access to active data and enabling rendering processes to run at the high performance that ILP production requires. Jacobson says, “We use the combination of EC2 Spot Instances and a cloud-based Avere cluster to give us rapid deployment of extra computing power when needed.”

ILP is now in a position where it can take on new business secure in the knowledge that its infrastructure will cope. And artists can get on with creating stunning VFX without worrying about where their work will be rendered. Jacobson says, “Rendering on AWS gives us the freedom to pursue opportunities knowing that we can scale to deliver high-quality VFX on time and within budget.”

At the end of a project’s lifecycle as deadlines draw near, ILP can boot up additional cloud render nodes to stay on schedule. So far this has worked well in projects for big-name titles such as Childhood’s End on the Syfy Channel. “To give you one example, we completed our work on Childhood’s End at the same time as finishing season three of Black Sails, which was using all of our on-site resources. Doubling our permanent infrastructure for a week’s worth of work would not have been feasible, but by taking advantage of the AWS cloud, we were able to double the size of our render farm and deliver great effects while meeting the tight deadlines of two shows,” says Jacobson.

If it didn’t have its low-latency cloud render farm with AWS and Avere, the alternative would be to either invest in permanent hardware on site, or turn down work. “We never take on a job unless we can guarantee we’ll produce high-quality work, so it would be a case of making a large, up-front investment,” says Jacobson. “Working in the cloud is an exciting model for our business. It’s a flexible way of buying compute resources, not least because it’s a clean way to tie costs to a specific project. Using Spot Instances really makes it worthwhile for us as well, because we can get a lot more machines for our money.”