HPC news
hPc helps deliver animated feature film
Framestore, the Oscar-winning peak working we were generating animation proceeded alongside ‘With this capacity we believe we
animation and computer 5TB per night, so reliable high the development of the are, by some margin, running the
generated imagery studio, has performance storage was central processing needed to render this most powerful CGI render farm
just completed its first full length to our system design. We were final look. This parallel approach in Europe.’
animated feature, The Tale of very keen to use RAID 6 to increased storage and render farm The storage cluster technology
Despereaux for Universal Studios, give us extra resilience. We demands, as many intermediate used is Lustre, the open source
using its in-house developed high selected Infortrend’s EonStor stages of production data had to Linux cluster file system, which
performance computing system storage systems as they gave us be retained to avoid unnecessary combined with the open
for the rendering of the images. the capacity and flexibility we animation rework whilst fine architecture of the Infortrend
This system was made up of needed. Having used Infortrend tuning of the lighting effects and storage arrays, and server
a 200TB clustered Linux file products for the past 10 years we colour proceeded. hardware from HP and Dell,
system closely coupled to a 6,000 were confident we could trust ‘Eighty per cent of the provided Framestore with a
core render farm. Production them for this critical project,’ rendering for the film was scalable infrastructure. At the
data was mirrored on a second said Steve Prescott, technology completed in the last three peak of production the cluster
400TB cluster, all of the data director at Framestore. months of production. We was using 4,700 of its 6,000 cores
storage was based on Infortrend The creative team had been designed the systems to scale. for work on the film – with other
EonStor storage systems. briefed to achieve a unique ‘The project started with 50TB projects still running in parallel.
‘Producing cinema quality look for The Tale of Despereaux, of storage, which we grew by an The processor cores are split
animation generates huge inspired by the Dutch Masters. additional 150TB as we worked,’ across 1,000 Dell servers each
volumes of valuable data. At To meet the required deadlines said Prescott. He continued: core having 2 GB of RAM.
hPc Products
New copper cable assemblies providing reduced jitter, a very low With this new built-in library, from the addition of a Graphics
from Gore latency, and a maximum signal eye scientists, engineers and analysts Toolbox for improved visualisations.
among the new hPc-related opening. can process extremely large image The licensing and pricing model
products are two copper cable www.gore.com/electronics data sets and perform complicated is as similar as possible to the
assemblies from Gore. The QSfP imaging analysis on hPcs with matlab licensing model to make
copper cable assemblies are Star-P gets imaging boost great speed and ease. purchasing and maintenance easy.
ideally suited to Infiniband, 10 Interactive Supercomputing has Star-P enables users to create Jacket is used by customers
Gigabit ethernet and 40 Gigabit added powerful new imaging applications on their desktops using across all major hPc industries,
ethernet applications and can capabilities to its Star-P technical familiar mathematical tools – such such as the automotive, financial,
improve signal reliability at speeds computing software, boosting as matlab and Python – and then medical, and seismic industries.
over the standard Infiniband QDr productivity by executing functions run them instantly and interactively In order for high performance
data rate 10 Gbps/channel. at the up to 30 times faster than on parallel hPcs. Problems that computing (hPc) companies to
same time, Gore’s design and the competing tools. can take hours to run in matlab or adopt GPu technologies, a robust
low dielectric materials used in the The Star-P imaging function other environments on the desktop and healthy software tool chain
cable’s construction improves bend library was developed with the help can often be run in minutes with must be created to connect
radius over longer distances in of anona Labs, an Israeli software Star-P’s image processing function programmers to GPu hardware.
small cross-sectional areas. company specialising in advanced library. While hardware manufacturers are
also available are the SfP+ image processing technologies www.interactivesupercomputing. building lower-level software tools,
copper cable assemblies for the for interactive numerical analysis com such as cuDa, which support their
hPc, enterprise networking and environments. The library is ideal devices, accelereyes delivers high-
network storage markets. These for use in life sciences, defence Jacket 1.0 launched level interfaces which remove the
cable assemblies offer a more and any other application involving accelereyes has released Jacket lower-level complexity.
reliable, lower power and more complex imaging problems with v1.0, its GPu engine for matlab, Jacket’s Graphics Toolbox
cost-effective alternative to fibre high computational requirements marking the end of a seven month enables true Visual computing,
optic modules for transmitting 10 or with extremely large data beta release cycle. Jacket now seamlessly merging the compute
Gbps data up to 25m. Gore SfP+ sources, such as medical imaging, supports a large number of matlab power of cuDa with OpenGL
copper cable assemblies comply microscopy, surveillance and functions, each offering great visualisations.
with the Sff-8431 standard, while biometrics. speed, and has also benefited www.accelereyes.com
28
SCIENTIFIC COMPUTING WORLD february/march 2009 www.scientific-computing.com
SCWfeb09 pp26-29 HPC news.indd 28 4/2/09 10:44:11
Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44