Chrysler Adopts Linux For Vehicle Simulations

eMilkshake writes "According to this ComputerWorld article, Chrysler is adopting Linux for vehicle crash testing. According to the article, 'the new system is expected to improve simulation performance by 20%, while saving about 40% in costs....'" Insert knee-jerk reaction joke about computers and crashing here.

computers

[Anonymous Coward]

crash

RTFA!

[scrod98]

"The cluster is running modified LS-DYNA crash-testing software from Livermore Software Technology Group in Livermore, Calif."

It's like, right at the beginning.

Other references

[mmol_6453]

eWeek [eweek.com]
Computer Graphics World [pennnet.com]
Business Week [businessweek.com]
Globetechnology.com [globetechnology.com]
ZDNet [com.com]

The wonders of news.google.com [google.com].

Re:Software?

[rhombic]

Read the article? "The cluster is running modified LS-DYNA crash-testing software from Livermore Software Technology Group in Livermore, Calif." Now if we could just get moderators to read the article before going +1 informative....

Re:Wait a minute

[Osty]

So wait a minute, you save money by using Linux? Are you telling me that the folks at MS are lying and that the TCO for Winodws isn't lower?

Red herring. The article was about a migration to linux from a proprietary unix, not from Windows. Therefore, the only conclusion you can draw here is that linux saves you money over other unix operating systems. You can't make any conclusions about linux TCO vs. Windows TCO based on this article.

Care to try again?

Re:This is one application -- You don't understand

[Proudrooster]

The application itself if not going to run on the cluster. The cluster is simply going to be used as a "compute farm" for solving the datasets and models produced by the application.

This is similar to what is happening in the animation industry. The LINUX boxes are simply going to "crunch" the numbers and feed the results back to an application running under Windows or high end UNIX workstations.

For a cheap "compute farm" cluster, you can't beat Red Hat Advanced Server with Xenon's.

We are planning to build a 16 node cluster next year for the same purpose as Chrysler. Again, the apps aren't running here, LS-Dyna, DynaForm, Hypermesh, FEMB etc ... will still all run on a UNIX/Windows workstation, but the solving will be done (very quickly) on a Red Hat cluster.

It just rocks!

Re:Multiple Crashes.

[joib]

Chrysler is a big company, so they certainly have many projects going on at the same time. By centralizing on one big cluster they can increase utilization compared to each department needing simulation capabilities having its own minicluster. So what the article probably means is that at the same time they can run 18 different simulations each using on average 12 cpus (108*2/18).

It's the same on (almost) all supercomputers. They have lots of users, but most users don't use that many cpus for their jobs. Take me, for example. On the supercomputer where I have an account there are 512 power4 cpus. Usually I use 8 or 16 cpus for my simulations, and so do almost all the other people using the same machine. About the only time the entire supercomputer is reserved for one job is when they're benchmarking it, which as you certainly can imagine isn't done so often.