Linux Cluster Makes Impact on Chryslers Bottom Line

Linux cluster based on IBM workstations will perform impact analysis simulation at faster speeds and a lower cost.

DaimlerChrysler has found an innovative way to cut costs: It is using a Linux cluster based on IBM workstations to perform impact analysis simulation at faster speeds and a lower cost.

An IBM spokesman told eWEEK that the Chrysler Group, a unit of DaimlerChrysler, is the first automotive company to take advantage of the price of a Linux-based workstation to obtain the power of a supercomputer at a cost of 40 percent less than other solutions available on the market.

This new Linux cluster, which IBM and the Chrysler Group will detail on Monday, is based on 108 IBM IntelliStation M Pro 6850 workstations powered by dual Intel Xeon processors operating at 2.2GHz each with enhanced 512KB L2 cache.

Intels 2.2GHz Xeon processors and PRO/1000 server adapters are used for high-level performance and flexible connectivity, and the solution includes TotalStorage FAStT500 using Gigabit Ethernet.

Livermore Software Technology Corp modified its LS-DYNA software--to comply with Linux standards--to perform the crash test simulations, while Red Hat Inc. provided the Linux operating system.

Vehicle and occupant impact simulation accounts for some 70 percent of the Chrysler engineering groups total computing capacity, so the need for new and innovative approaches that do not compromise design creativity is significant, said Bernard Robertson, senior vice president of engineering technologies and regulatory affairs for the Chrysler Group.

The shift taking place in the price/performance curve on impact analysis technology "is empowering our engineers to develop future vehicles with better decision-making capability and at a rate that is faster than before. We expect to see a direct correlation to the quality and safety measurements of our vehicles, as well as improved productivity at less the cost," he said.

In the early 90s, the Chrysler Group used one supercomputer, which cost tens of millions of dollars, to run tests in a virtual world. Then, between 1996 and 1999, it moved to a network, or cluster, of Unix-based computers for more efficient simulation computing, at a faster rate and cheaper cost, Robertson said.

The Chrysler Group will continue to look at new ways of applying the latest Linux cluster technology to its simulation strategy for computer-aided engineering, he said.