Building the Death Star as a challenge for NetApp

And NetApp participated in the construction of the Death Star. :)
Deathstar - “The Death Star” is the historical name for the computational cluster used by the 3D modeling division of LucasFilm, ILM company - Industrial Light and Magic.

Since 2007, NetApp storage systems have been used to store data used by the entire LucasFilm IT complex, which includes ILM (Industrial Light and Magic), their 3D modeling and rendering studio, video production and editing, the LucasArt division (video games), and the Starwars website. com.
Storage capacity today exceeds petabyte of data.


The applied technology of work in film production requires access to a huge array of common, accessible to many different data consumers. So, for example, in the case of a 3D render farm, these will be models and textures, calculated for each frame in parallel by many hundreds, and even thousands of processors, in the case of nonlinear video editing, different fragments of finished scenes. Finally, the website starwars.com, which stores a huge amount of diverse material about the Starwars Universe, serves many tens and hundreds of thousands of hits from around the world to LucasFilm webservers daily.
')
Traditionally, to solve this problem, a distributed multi-node cluster operating as an NFS server is used. Lucasfilm has been using NetApp storage systems for many years (since 2007, and before this time - Spinnaker, acquired by NetApp in 2004), and currently only the storage capacity of various levels exceeds petabyte (this does not include archival, offline -store on magnetic tapes).
NetApp controllers work as nodes of a parallel NFS cluster running OS Data ONTAP GX, allowing you to create a single “storage field”, a single common “namespace” for stored files from all these individual physical nodes, and ensure high performance data access that nowadays 3D-video with ever-increasing demands for quality, number of CG (Computer Generated) characters and scenes, and production volume, is one of the most important tasks for a film production company.

I have already told in this blog about a similar repository of another large company working in the field of 3D rendering and video production - New Zealand's Weta Digital (some more information can be found in blog.aboutnetapp.ru blog, in articles here and here ), in which there are also many The NetApp repository has been used for years for her numerous film works, such as the Lord of the Rings trilogy, King Kong, District9, Avatar, and many others.

Back in 2004, LucasFilm built for its data center a network backbone (currently the largest in the world) operating at 10Gb / s, using Brocade solutions, and providing data to a huge render farm (second in size and power in the movie world after Weta), today using IBM blades (older - LS22 on AMD, newer - HS22, on Intel Xeon, 32GB RAM per blade).

Today, the main data center LucasFilm occupies about 1,000 square meters of space, where the main manufacturing computing facilities of the company are located in San Francisco (there is also a LucasArt division in Singapore engaged in video and computer games).

Looking at such huge computing resources, it seems that the growth possibilities of the data center are endless and you can ignore its costs, but this is not so. “Our possibilities, unfortunately, are always strictly limited by the film’s budget,” says Peter Hrichek, LucasFilm

All that the data center spends in the form of side costs, such as, for example, cooling, power, is taken away from the productive possibilities, from what the company can spend on the film itself. Therefore, issues of efficiency of use, that is, the ratio of how much we get to the budget dollar spent, despite the seemingly huge financial capabilities of such companies, are very serious. In order to give an idea of ​​the costs of such projects, we can say that the data published in 2007 show that the LucasFilm data center consumed at that time about 2.4 Megawatts of electrical power, of which only about half directly fed the server, the storage system and network hardware. For electricity prices in San Francisco, California, this is a very significant expense.
For comparison, for example, this consumption of electricity is about one and a half thousand average American houses. One data center consumes electricity as a whole small city!

As for storage, it continues to grow, following the film's increasing budgets. So, according to the published data, the operational storage capacity of the Lucasfilm data center was only 32TB in 2004, increased to 300TB by 2007, and today amounts to more than 1 petabyte (1000 terabytes), three-quarters of which are high-performance storage, and the rest is nearline, there is online data with lower performance, for example, on SATA disks.

A short video tour of the data center Lucasfilm was recently published by the journalist Arik Hesseldahl in the online publication All Things Digital.

http://newenterprise.allthingsd.com/20110325/lucasfilms-data-center-and-an-encounter-with-the-real-death-star-video/

Unfortunately, I can not insert this video here, go to the site by reference. The video, by the way, is accompanied by English subtitles, which is very convenient for those who can’t understand English badly.