Google Cloud Platform Blog
Product updates, customer stories, and tips and tricks on Google Cloud Platform
Framestore Frees Up Designers to Create Unforgettable Visual Effects, with Help from Google Compute Engine
November 19, 2014
Today’s guest blog comes from Steve MacPherson, chief technology officer for
Framestore
, a visual effects production firm headquartered in London. Framestore’s visual effects work has been seen in films like “Avatar,” “Gravity,” and “Guardians of the Galaxy,” winning the company numerous Academy Awards, BAFTAs, and Cannes Lions awards.
If you saw “Gravity,” hopefully you enjoyed the movie’s depiction of space travel and weightlessness. It was a brilliant experience to take on such a daunting project, and very rewarding to be part of the collective effort.
A lot of planning and effort goes into every visual effect, as well as a fairly large amount of computing power. At peak times – when we’re rendering images on several projects at once for advertisers and film studios – we’ll consume the processing power of up to 15,000 Intel cores. Managing peak provisioning and matching resources to projects is central to keeping the production pipeline moving toward various deliveries. The challenge that returns regularly is when demand for resources conflicts with capacity – usually during periods when the stress of delivery is at its highest, and the focus is on realizing the creative goals of our clients in time for the immovable object that is a major release date.
Historically, this boiled down to a simple scenario: purchase additional equipment. A design maxim I've long held is that we've never built a machine room that doesn't eventually run out of space, cooling or power. This “computational load bubble” is a result of both the scale of modern studio films, and the fact that we run multiple films through the facility in parallel. This is the peak provisioning problem that we were looking to address for a number of years. Once films are delivered, the demands recede, and we have an excess capacity for some period of time until we reach the next set of deadlines.
For the past few years, we’ve kept a close eye on the potential for using external resources as an overflow valve. Google, through its
Google Compute Engine
, is the first company we've worked with that was able to combine raw resources with a team that understood our requirements in detail and a business model that helped us manage the economics. Within a day of firing up the network, we built our image inside the Compute Engine container.
At Framestore, we’ve developed a sophisticated in-house job submission system based on our render queue manager, fQ – it’s extremely efficient at juggling various job types to match them accurately to rendering nodes available at any given time. This workflow is central to the sustained high levels of efficiency for our render farm, giving us up to 95% of overall capacity for weeks on end.
Having Google Compute Engine on the back end opened a number of opportunities for us to siphon off a certain class of work during a period of peak production. The load reduction on our farm allowed us to be much more specific in how we prioritized our deliverables, ultimately leading to a much more focused and predictable delivery schedule – great for production and great for maintaining confidence with the studios.
Google gives us breathing room during periods of peak capacity, allowing our artists more flexibility around creative refinement. We do many iterations of an image or a visual effect, making minor technical tweaks and submitting it to the render farm to see if it works. During periods of high load, if all of our rendering is in-house, the creative team might have to wait more than a day to see results when the in-house farm is at capacity. This introduces stress and management overhead around which shots get priority.
By adding Compute Engine to our workflow and allowing our in-house capacity to focus on the studio work, everyone’s project gets computing time – and the creative team can get as imaginative as they want to, with fast views of new iterations.
The results: fewer bottlenecks, more creativity and more predictability, not to mention saving about £200,000 (more than $300,000 USD) on the cores we didn’t need to buy. We can now confidently move into final stages of production on our biggest projects, knowing we have a reserve of computational ability on tap. When you check out new movies this year and next like “Dracula Untold” and “Jupiter Ascending,” you’ll be looking at our visual effects work, created using all the computing power at our fingertips.
No comments :
Post a Comment
Free Trial
Labels
Android
Announcement
api
app engine
Atmosphere Live
bigquery
BigTable
CDN
Cloud Console
Cloud Dataflow
Cloud Datastore
cloud endpoints
Cloud Pub/Sub
Cloud SDK
cloud sql
cloud storage
Cloudera
Compute
Compute Engine
container cluster
customer
Dev Tools
developer tools
developer-insights
Developers
Developers Console
devfests
Disaster Recovery
Encryption Keys
ESG
Event
events
GA
Go Client
Google App Engine
Google Apps
Google BigQuery
Google Cloud Deployment Manager
Google Cloud Networking
Google Cloud Platform
Google Cloud Storage
Google Compute Engine
Google Container Engine
gRPC
hadoop
Hardware
Helium
how to
IO2013
iOS
Kubernetes
Levyx
Local SSD
mapreduce
Media
Nearline
networking
open source
PaaS Solution
Partner
Pricing
Research
round-up
Server
Siggraph
solutions
Startup
Tableau
TCO
Technical
Windows
Wowza
Zync
Archive
2015
Nov
Oct
Sep
Aug
Jul
Jun
May
Apr
Mar
Feb
Jan
2014
Dec
Nov
Oct
Sep
Aug
Jul
Jun
May
Apr
Mar
Feb
Jan
2013
Dec
Nov
Oct
Sep
Aug
Jul
Jun
May
Apr
Mar
Feb
Jan
2012
Dec
Nov
Oct
Sep
Aug
Jul
Jun
May
Apr
Mar
Feb
Jan
2011
Dec
Nov
Oct
Sep
Aug
Jul
Jun
May
Apr
Mar
Feb
Jan
2010
Dec
Oct
Sep
Aug
Jul
Jun
May
Apr
Mar
Feb
Jan
2009
Dec
Nov
Oct
Sep
Aug
Jul
Jun
May
Apr
Mar
Feb
Jan
2008
Dec
Nov
Oct
Sep
Aug
Jul
Jun
May
Apr
Feed
Technical questions? Check us out on
Stack Overflow
.
Subscribe to
our monthly newsletter
.
Follow @googlecloud
No comments :
Post a Comment