The largest commercial Linux installation

by Dhiram Shah

Almost all movie studios primarily use Linux for animation and visual effects, but with more than 1000 Linux desktops and 3000 Server CPU’s Dreamworks Animation is the largest commercial Linux installation. We all know Moore’s law which states that computing power will double every two and a half years, here is a little known corollary from the animation world ‘cartoon animation CPU render hours will double every three years’. In 2001, the original Shrek movie used about 5 million CPU render hours. In 2004, Shrek 2 used more than 10 million CPU render hours. And in 2007, Shrek 3 used 20 million CPU render hours. Dreamworks uses the popular Linux package Maya for animation. Solid support for threading, NFS and LAMP toolsets are a few of the advantages with Linux. At the desktop, Dreamworks uses HP xw9300 workstations running RHEL 4 and the renderfarm uses HP DL145 G2 servers, with 2GB per core the servers have 8GB of RAM as they have 4 cores.

DreamWorks uses its own TIFF-like file format based on 16-bit binary fixed point, a limited High Dynamic Range (HDR) image format with a color range of 0 to 2.0. Letting images go whiter than white leaves headroom for image adjustments.


  1. Monkey

    “1000 Linux desktops and 3000 Server CPU’s Dreamworks Animation is the largest commercial Linux installation.”
    Depends on what you consider commercial. Here are a few points about my data center.
    1) Google has 20,000 servers (confirmed)
    2) Google is in 8 of my data centers (8×20,000 confirmed)
    3) Google has those 8 mirrored (8×20,000×2)
    I can personally confirm the 20,000 in my data center. Via fellow co-workers in the other data centers (8×20,000) = 160,000 is confirmed. Via conversations with the google techs; those data centers are mirrored with other providers and that is 360,000 servers.

  2. Anonymous

    This is actually FAR from being the largest commercial linux installation. In the HPC world there are numerous companies with very large installations, most of which will not publish their compute power numbers for competitive reasons. One company in particular has well over 10,000 dual processor, dual core nodes (40K processors) in just one of several of their compute centers.
    Beyond that, IBM and HP both run very large Linux clusters for compute-on-demand services, and Google’s compute farms are composed of thousands of Linux based systems (though not not a high-performance compute infrastructure)


Leave a comment