Public

The aim of the project, which is financed equally by Airbus, Lower Saxony and DLR, is to develop processes, methods and numerical computing procedures that will enable extremely accurate simulations of the overall flying behavior of an aircraft, even in the very early stages of its development. C²A²S²E will allow the technological, economical and ecological risks of the air traffic, such as noise pollution and emissions, to be crucially reduced.

T-Systems delivered the platform that makes it possible for the C²A²S²E project to carry out complicated numerical flow simulations. Dr. Alfred Geiger, head of Solutions & Innovations Scientific & Technical ICTICT

Information and Communication Technology. These two terms began to be combined in the 1980s to emphasize the growing convergence between the two technologies.

at T-Systems Solutions for Research GmbH:

“We were able to offer a complete service that meets the requirements of industrial developers in design and optimisation, as well as the needs of public research in terms of feasibility and verification of technical solutions. This was only possible based on our longstanding cooperation with DLR, and our experience with the hww public private partnership (hww - Höchstleistungsrechner für Wissenschaft und Wirtschaft GmbH)”.

Alongside incorporating user requirements, the system had to be embedded in DLR‘s infrastructure at their research airport in Braunschweig, and connected to the various security architectures of the partners, Airbus and DLR.

From the project launch and solution design right through to production, the project took just over nine months. „Considering that the latest high-end technologies were used across the project, this is outstanding,“ said Dr. Geiger. Today, high-performance computing ( HPCHPC (=High Performance Computing)

High Performance Computing refers to the use of parallel supercomputers and computer clusters to perform complex computations.

) no longer means simply purchasing a complete solution from one manufacturer but, instead, is the project-style implementation of requirements in close cooperation with end users, service providers and technology suppliers. For example, the hardware has to be adapted to the software – not the other way around – as application software (TAU CFD code) and data management lifecycles are much longer compared with hardware lifecycles.

When it opens in May 2008, the C²A²S²E-simulation center will be one of the top 40 HPC systems in the world. Even more, it is the second fastest computer worldwide for industrial use which is not installed at a computer manufacturer.

The deployed Petascale architecture of Sun Microsystems is the first completely non-blocking and equidistant cluster architecture in the world that in a fullblown configuration allows processing speeds of over 2 petaflop/s (2 quadrillion floating point operations per second.) The C²A²S²E blade servers use AMD Opteron processors as compute nodes, which are operated using modularly structured Sun system chassis. The primary component for interprocess-communication is the Sun Data Center Switch 3456; the world‘s only InfiniBand Switch with 3,456 ports. The switch‘s high packing density enables data throughput with minimum latency, which is essential for high-performance compute systems. This maximum data throughput allows the simultaneous computation of several highly complex simulations in throughput mode as well as the use of the overall system for a particular high end application. This creates crucial added value for the research and development community.

„It was important to us to significantly increase the number of processor cores in use,“ explained Prof. Norbert Kroll, Head of C²A²S²E Center at the Institute of Aerodynamics and Flow Technology, at DLR in Braunschweig. „We were using on average 100 to 500 cores at a time, but we intend to gradually increase this figure to 6,000 cores.“

The deployed 768 Sun Blade Server modules actually make 6,144 cores available to the users. The modules each use two AMD Opteron Quad-Core processors and are supported in terms of storage by Sun Fire X4500 and 6140 servers and at the front end by Sun Fire X4200 M2 servers. To ensure the server technology remains state of the art, the blades will be replaced in 2010 with the latest technology as part of a refresh program.

Operating a powerful cluster system of this type is simply not possible without efficient, high-performance air-conditioning. Two redundant refrigeration units with a total cooling power of 900 KW provide the cluster system‘s 12 cooling units with the necessary cooling water via a complex system of pipes and two buffer storage units, each with a capacity of 4,000 liters.