High Performance Computing Center Stuttgart

Ein state-wide Ressourcen-Grid in Baden-Württemberg

BMBF funding for infrastructure led to the emergence of the bwGrid from the InGrid project. In bwGRID, the functionality and value of grid concepts in the HPC environment are proven and previous organisational and security problems are overcome. These problems can be primarily attributed to distributed data management, the various types of user administration or the lack of a security infrastructure between different computing centres. Furthermore, new cluster and grid applications are being developed, the licensing problems are being dealt with and it will also be made possible for individual computing centres to become specialised. Each computing centre will dedicate itself to one key area and coordinate support for the users. Each centre will support one other grid computing focal point (that is as closely related as possible) in order to link up local and distributed competencies.
Storage concept.

The network of the state-wide grid for Baden-Württemberg is rounded off by a cross-state storage solution that consists of local components in the individual cluster locations and a central storage system. The system as a whole is referred to as federated storage and is administered and used across organisational boundaries. Naturally, a distributed storage concept such as this also brings challenges with it. On one hand, the reliability and stability of the global file system is extremely important since it is only possible to use the computers in the bwGrid on a restricted basis if the central file system is not working. On the other hand, the proposed interaction of local and central storage components in different facilities is still uncharted territory. There is a particular need to train users in how to handle the new systems and to explain the underlying concepts to them.

Hardware details

IBM blade server: 2x Intel Xeon Quadcore 2.8 GHz, 16 GB RAM.
Freiburg, Heidelberg, Karlsruhe, Mannheim and Tübingen each have 10 BladeCenters (140 nodes). These clusters each have a computing power of 9.59 TFLOPS, giving them places 402 to 406 in the list (June 2008) of the world’s fastest computers.
20 BladeCenters (280 nodes) in Ulm with a computing power of 19.18 TFLOPS, putting it in 98th place.
31 BladeCenters (428 nodes) in Stuttgart with a computing power of 29.32 TFLOPS, putting it in 65th place in the TOP500.
Some of the computing centres are already connected with each other at a speed of 10 GBit/s via Baden Württemberg’s extended LAN (BelWü).
The parallel computer clusters procured for bwGRID will be coupled to a network as part of the D¬-Grid Initiative.

Automotive simulations
Founded in March 2008, the Automotive Simulation Center Stuttgart
(ASCS) will ensure that the fields of research and industry cooperate even more closely in vehicle construction. The new centre focuses, among other things, on lightweight construction on the development of vehicles that are quieter and, above all, consume less fuel and emit fewer pollutants. Numerical simulation methods for product development and optimisation are essential for these tasks as well as others in the automotive sector.

Industry participation
In order to ensure that the developments from bwGRID are sustainable, use of the infrastructure will be promoted in industry and areas close to industry in addition to the purely academic world.

However, this is often accompanied by considerably higher requirements in terms of data security and availability than those of academic users. This orientation to the requirements of industry can also create cost transparency which can promote a greater awareness of costs.

Climate simulation
The aim behind cooperating with the German Weather Service (DWD) is to improve the simulation of meteorological phenomena. To this end, efficient, higher order procedures will be developed and validated on unstructured grids in 3D for use in meteorological simulations. One aim of the project is to enhance the mathematical model so that different weather phenomena (e.g. rain, ice) can be taken into consideration. The work group for trace element modelling and climate processes at the Karlsruhe Institute of Technology (KIT) is using computer-aided simulations to investigate the spread of gaseous and particle trace elements in the troposphere. The work groups’ focus here is on analysing the processes that are significant to the distribution of trace elements in terms of time and coverage. Aerosols play a key role in the climate system. They alter the radiation field and react with clouds and gaseous trace elements.

Simulation of the temperature ranges of the cosmic microwave background radiation for a 3D torus universe by using the first 5.5 million eigenfunctions (‘overtones’). The oval map shows the entire sky (in a Mollweide projection). The different colours denote hot (yellow) and colder (blue) temperatures.

The research area of ‘old age provision and savings behaviour’ analyses the existing social security systems and the interaction between them and private savings. The effects of demographic change on the financial viability and organisation of the pension, health and nursing care insurance systems are studied and researched as well as how and why people save.

Energy from fuel cells
The ‘Model-based design of fuel cells and fuel cell systems’ research network is embedded in the BMBF framework for fundamental research into renewable energies.
Mathematical models are developed and used to enable computer-based design of PEM (polymer electrolyte membrane) fuel cells, stacks and systems. On this basis, detailed 3D simulations of fuel cell stacks are carried out by using a complex model that describes the fuel cells as closely as possible. The results of these simulations are compared with measurements taken from prototype fuel cells.

Life sciences / chemistry
The combination of biological and chemical methods in various scales of length and time, ranging from quantum chemical ab initio methods to molecular dynamics simulations and thermodynamic/biological models, makes it possible to examine highly complex systems. For example, during the design of active pharmaceutical substances, it is possible to test the suitability of a number of potential substances on a computer within a relatively short space of time (screening). Only the ingredients that are shown to be highly suitable by the test undergo further, more detailed experimentation. These and similar procedures can significantly increase the efficiency of designing new substances.