The first mainframe system at Audi for CAE applications was used in the late eighties of the past century. It was followed by a Cray Y-MP2E supercomputer with 512MB Ram in the early nineties. Shortly afterwards there was the introduction of RISC based SMP servers. In the late nineties Audi scored in the TOP500 list with the Fujitsu VPP300

vector computer using 16 CPUs. With the turn of the century came Audi's first Linux Cluster, as the speaker showed the audience in a timeline of past events.

Audi deployed the first Linux cluster of servers in September 2001. In June 2002 the first Linux workstations were deployed and since mid 2003 Linux has been the strategic server and workstation platform for CAE at Audi, as Mr. von Bülow explained.

Since mid 2003 Audi deploys more Linux than UNIX servers. By the end of 2007 more than 90% of all server

resources will be 64-Bit (x86_64) Linux based. In addition, since mid 2005 Audi is using more Linux than UNIX

workstations. The speaker predicted that by the end of 2007 more than 80% of all workstations will be 64-Bit (x86_64) Linux based.

Mr. von Bülow also stated that x86-64 with Linux will continue to be the platform of choice for Audi. This does not mean that there are no problems with Linux. On the contrary. The speaker explained that Audi uses Catia V4 and V5 from Dassault Systèmes and Pro/ENGINEER (PTC) as the strategic tools for computer-aided design (CAD). Not all Pro/ENGINEER modules however are available on Linux, e.g. Pro/MECHANICA.

In addition, Catia V4 is, due to its age, available only on legacy UNIX systems. But worst of all is that Dassault Systèmes has until now refused to release a Linux version of Catia V5, according to the speaker.

Mr. von Bülow went on to address the consequences and challenges that Audi faces in deploying a larger number of systems. When you dispose of a large number of systems this involves that an even larger number of simulations are carried out, resulting in a large amount of raw data being produced. For example, an average crash-simulation generates between 500 and 2000 MB of results. At 2000 simulations per day the CAE-engineers generate 1.2 TB of data per day, the speaker explained. Now what is Audi doing with all of that data?

In order to handle the large amounts of produced data Audi employs the Simulation Data Management (SDM) System CAE-Bench. The CAE-bench is a customised version of Virtual.Insight developed by MSC.Software.

According to the speaker, the CAE-Bench is able to automatically start simulations, store the produced data, and

generate reports with the key-results based on templates.

The next generation of CAE-Bench which is aptly named CAEBench II, will increase the capabilities for SDM. Mr.

von Bülow explained that the CAE-Bench II is a customised version of SIM.Manager, that is also developed by MSC.Software. The CAE-Bench II will greatly enhance the support for simulation work flows at Audi so that a reduction of the amount of customisation becomes a key goal.

As far as the future projects are concerned, Audi aims at bridging the gap between CAD and CAE by coupling its

SDM system with the corresponding CAD data management systems. This will enable the automation of the CAE model generation which has the potential to dramatically shorten Audi's design cycles and increase the need for HPC solutions, the speaker stated. This coupling will happen within the next 1 or 2 years.

In the long run, Audi plans to Grid-enable the SDM system in order to encompass distributed locations, multiple organisations and the integration of design, test and simulation tasks. The EU-Project "SIMDAT" is currently working on this enablement. A partial implementation of the developed techniques is anticipated over the next 5 years, as Mr. von Bülow announced.

He concluded by saying that the throughput is more important than the peak performance at Audi and hoped that the CATIA V5 for Linux would be coming soon. In general, Mr. von Bülow was convinced that the biggest challenge in the years ahead will not be the question of who will be having the biggest computer but of who will use the best strategy for handling the produced data.