Bold Visions Become Reality
Translation of the folder "Kühne Visionen werden Realität"
Science on the Highest Level:
High Tech in High Speed Quality
A milestone on the path to an information- and science-based society
High performance super computing is one of the most pivotal aspects of successful information technology. Whether in genome research, fluid dynamics, or climate analysis, super computers play a decisive role today in numerical simulations --- the third pillar beside theory and experiments in technical and scientific discovery. High performance super computers are necessary not only in academic research but also in industrial applications, such as space exploration or automobile manufacturing, in order to thrive in global competition.
With the expansion of the Leibniz Computing Centre (LRZ) of the Bavarian Academy of Sciences to a national centre for high performance supercomputing, Bavaria once again underscores that it is on the leading edge in development and application of innovative technologies and sciences. Significant funds from the privatization efforts have made it possible to realize this project in a very short time frame. Bavaria therefore remains an excellent location for science and research.
Like no other state, Bavaria combines high performance supercomputing and compatible pre- and post- processing computers at large universities. Moreover, Bavaria is a pioneer in the development of Gigabit technology and has at its disposal extremely fast broadband data networks. Furthermore, Bavaria sets as high priority the expansion of its universities: The expenditures for education, science, and research in Bavaria are above average levels. In fact, at 3% of gross national product, the research expenditures are world class. These efforts pay off well: Approximately 1/3 of all German jobs in the computer industry and almost 40% in the field of electronic media can be found in Bavaria. Furthermore, the Munich region is already in second place Europe-wide in the fields of information and communication technology. Holding onto and expanding upon this regional advantage is an important priority for science, economics, and politics in Bavaria. The new SR8000-F1 of the LRZ is in this sense a further and necessary step in the right direction. Resolute and responsibly used high technology creates security and the future!
The New High Performance Super Computer at the Leibniz Computing Centre
Since March 2000, the first Teraflops-Computer in Germany has been in operation at the Leibniz Computing Centre of the Bavarian Academy of Sciences in Munich. Already in the first delivery phase, the new Hitachi SR8000-F1, with a top computing capacity of 1.34 Tflop/s (1 Teraflop/s = 1 trillion FLoating point Operations Per Second) is not only the fastest computer in Germany; at the time of installation it is also the fastest civilian computer in the world. In a second installation phase, the computer will be expanded to 2.2 Tflop/s. In comparison, the highest performance computer in Europe up until now has been the .98 Tflop/s computer of the German weather forecasting office in Offenbach.
Modern cutting edge technology has its price: The new computer cost 90 million marks, including the costs for operation and service. Two thirds of these funds are provided by the free state of Bavaria and one third by the federal government.
What do we need a high performance supercomputer for?
High performance supercomputers will always be used when "normal" computers are no longer capable of fulfilling the enormous demands for computing power and storage space that are necessary for realistic simulations and modelling of natural phenomena and technical procedures. In this way for example in the field of medicine, operations can be supported with imaging techniques, or air turbulence structures can be investigated in the field of aeronautics. Crash simulations can be conducted in order to build increasingly safer automobiles, and weather forecasts can be improved through regional climate models.
For cutting edge research, extremely capable computers are critical today in all technical and scientific disciplines. As was already suggested in 1995 by the German Science Council (Wissenschaftsrat), the launch of the supercomputer Hitachi SR8000-F1 transforms the LRZ as planned into one of the centres for high performance computing in Germany.
Bold Visions Become Reality: The most powerful civilian supercomputer in the world
The computer system SR8000-F1 recently delivered by the Japanese company Hitachi is the fastest computer in civilian use in the world. Grouped in multiples of nine, its processors are operated as nodes with extremely high performance or they can also be used as single elements of a parallel computer -- a new type of supercomputer that unites the benefits of the old vector computers with the features of a modern parallel computer. The Hitachi SR8000-F1 at the LRZ currently has 112 computing nodes; at its final size it will have 168. In a computing node, eight of its nine RISC-processors, each with 1.5 GFlops peak processing capacity, can be combined through hardware and software to form a virtual vector CPU with 12 GFlops peak processing capacity. Therefore, a combined capacity of 1.34 TFlops can currently be reached, with an ultimate capacity of 2.2 TFlops (1 Tera-Flops = 1 trillion FLoating point Operations Per Second). With a traditional vector CPU, vectorizable computing operations are processed in a vector pipe so that per clock cycle one or more data words (i.e., numbers) are delivered to the CPU.
However, in the Hitachi SR8000-F1 a vectorizable computing operation is distributed over eight of the nine processors of a node (autoparallelization), and memory words are loaded into registers long before they are used. Therefore, all components of the eight processors can be used, not only the multiple floating point pipes as is the case in the more traditional vector CPUs.
Further performance features (final size data in parentheses):
- With a 928 (1376) Gigabyte main memory, even the largest jobs that have been proposed to date by users can be computed (1 Gigabyte = 1 billion bytes)
- With a total disk capacity of 7.4 (10.0) Terabytes, the resulting gigantic data amounts can be stored or read with large bandwidths (1 Terabyte = 1 trillion bytes)
- In the internal network, up to 86 (114) Gigabytes/s of data can be exchanged between the computing nodes (1 Gigabyte/s = 1 billion bytes per second)
- The nine processors of a node have access to a shared main memory
- The communication connection to the Gigabit-Scientific Network (G-WiN) of the DFN Association (DFN = Deutsches Forschungsnetz) and therefore the worldwide internet is realized through a HiPPI-interface, a fast data transfer method, with a theoretical bandwidth of 800 megabits/s (1 megabit/s = 1 million bits per second).
Not only the speed and performance of the new supercomputer are impressive, but also its physical dimensions, energy consumption, and dissipated heat far exceed those levels typically found at the LRZ:
- Dimensions (without disk array): 9m x 8m, 1.8m high
- Weight: 23 tonnes
- Energy consumption: 610,000 Watts
- Energy consumption of the total system (including air conditioning): almost 1,000,000 Watts (1 Megawatt)
Who can use the new supercomputer?
The new computer is available for processing computationally intensive as well as I/O intensive programs in research, science, and economics, only if those jobs cannot be computed on smaller systems. This means that researchers from all over Germany can process their jobs here. These jobs are orders of magnitude bigger and more complicated than was ever possible before. Jobs like these can be found in many research areas, e.g.:
- Astronomy: Nuclear combustion in stars, hydrodynamics and radiation transport in star atmospheres
- Biochemistry: Molecular dynamics in cell aggregates
- Chemistry: Structure determination of molecules
- Electrical engineering: Optimization of high voltage structures
- Geophysics: Dynamics of seismic waves
- Computer Science: Software engineering, methods for parallel applications
- Engineering Sciences: Coupled structural and fluid dynamics simulations, aerodynamics of automobiles and airplanes, crash simulations
- Mathematics: Adaptive procedures for solving differential equations
- Medicine: Imaging techniques for supporting operations and whole body modelling
- Meteorology: Regional climate modeling
- Physics: Simulations of high temperature supra conductors, simulations of melting processes and crystal growth
- Fluid Dynamics: Investigations of turbulence
The Steering Committee
The project proposals for resource allocation of the High Performance Supercomputer which are sent to the LRZ are forwarded to the Steering Committee. The Steering Committee consists of members of the Bavarian State Ministry for Science, Research, and the Arts, the German Research Council (DFG), and Bavarian scientists. Its twelve members judge the scientific claims of the proposal and the competence of the proposing scientists. They assure that the proposed projects cannot be computed with smaller computers of the hierarchical supply structure (work stations, group computers, compute servers in university computing centres, state high performance computers). They also ensure that the proposed projects are suitable for the Supercomputer, and evaluate whether the projects would be better suited to a supercomputer of a different architecture. For approved projects, the Steering Committee determines the duration of the project, the resource allocation, and possibly criteria for the prioritization of the project on the Supercomputer.
The Operations and Organizational Concept
The organization of the operation of the new supercomputer results from a collaboration between:
- The Leibniz Computing Centre of the Bavarian Academy of Sciences, as the operator of the new High Performance Supercomputer
- The Steering Committee of the High Performance Supercomputer
- The Competence Network for Technical and Scientific High- and Highest- Performance Supercomputing in Bavaria (Kompetenznetzwerk für Technisch-Wissenschaftliches Hoch- und Höchstleistungsrechnen in Bayern - KONWIHR).
The operation of the High Performance Supercomputer will be organized according to rules issued by the Steering Committee. The LRZ, as the operator of the High Performance Supercomputer in Bavaria, acts as the first source for consulting, especially with regard to those issues closest to the supercomputer itself and its operation. The LRZ also serves as a liaison between the users, the university departments, various institutions, and the Competence Network.
Since the following existing groups
- High Performance Computing in the User Consulting Department
- High Performance Systems in the Computing Systems Department, and
- Network Operations in the Communications Network Department
are already well prepared for the operation of the new High Performance Supercomputer because of their experience in operating the State high performance computers, new organizational structures were deliberately not introduced at the LRZ in order to take optimal advantage of synergies.
The Leibniz Computing Centre
The Leibniz Computing Centre (LRZ) of the Bavarian Academy of Sciences goes back to 1962, to the "Commission for Electronic Computing" (Kommission für elektronisches Rechnen) founded by the Academy members Hans Piloty and Robert Sauer. Now called the "Commission for Computer Science", it is still under the auspices of the Bavarian Academy of Sciences today. The commission, together with support from the Free State of Bavaria, successfully established a joint Computing Centre for Research and Education for all universities in Munich: What is today known as the Leibniz-Rechenzentrum, LRZ. Its name is a tribute to the great philosopher and universal scholar Gottfried Willhelm Leibniz (1646-1716), who reinvented the binary numerical system and invented one of the first working four-operation computing machines.
The LRZ, a modern service enterprise, is today:
- The scientific computing centre for the universities in Munich and for the Academy
- The centre for technical and scientific high performance supercomputing ("Supercomputing Centre"), and
- The competence centre for data communication networks.
The services of the LRZ can be used by all those affiliated with the Munich universities (approximately 80,000 students, 7,800 scientists, and 4,300 other employees).
The LRZ as University Computing Centre
"Computing Centre" is somewhat a misnomer: Computers today sit directly on the desktops. The foundation for the name is the concept of a "distributed, cooperative IT supply". It allows not only access to the worldwide internet, but above all it also provides constant flexible access to data and information, access to programs on other computers, the shared usage of expensive terminals, and much much more.
The task of coordination and liaison between the users is central. This task is achieved on a technical level through the operation of the Computer Network and the offer of Network services such as E-Mail or WWW and on a personnel level through courses and training programs. It is also achieved through software procurement and campus licenses, one-on-one consulting in problem cases, or development of new application possibilities.
The LRZ as Supercomputing Centre
For over ten years, the LRZ has operated the State High Performance Supercomputer for all Bavarian universities, with decentralized supplementing from compatible smaller development computers. Since March 2000, the LRZ is also operating the Hitachi SR8000-F1, a High Performance Supercomputer, which can be used nationwide. The LRZ operates not only the Supercomputers (SNI/Fujitsu VPP700/52, IBM SP2/77, CRAY T90/4), it also supports the conversion of appropriate algorithms onto the existing computer architectures. This happens in close collaboration with other Centres and Research Groups, especially with the Competence Network for Technical and Scientific High- and Highest- Performance Supercomputing in Bavaria (Kompetenznetzwerk für Technisch-Wissenschaftliches Hoch- und Höchstleistungsrechnen in Bayern - KONWIHR).
The LRZ as Competence Centre for Networks
The LRZ recognized the central importance of networks early on and has developed into a nationwide "Know-how" centre in this field. The newest communication technologies are piloted first at the LRZ, for instance the first X.25-2Mb-line, the first ATM-34Mb-line, the first 155Mb-connection, DQDB, and the first worldwide 2.5Gb-ATM-line (also with Gigabit-Technology with Wave Division Multiplexing (WDM)).
Research at the LRZ
The LRZ provides not only infrastructure and service provisions for users in science, economics, and technology; research also takes place at the LRZ. One of the priorities is modern network- and systems- management (Prof. Dr. Heinz-Gerd Hegering). The LRZ is also intensively integrated in research in the field of High Performance Supercomputing through its own work group, though the work with KONWIHR, and especially through the members of the Commission for Computer Science of the Bavarian Academy of Sciences: Prof. Dr. Roland Bulirsch (projectory computations of space ships, robotics), Prof. Dr. Karl-Heinz Hoffmann (mathematical modelling of modern materials) and Prof. Dr. Christoph Zenger (Permanent Secretary of the Commission for Computer Science and member of the Board of Directors of the LRZ).
Bavarian State Minister for Science, Research, and the Arts
Prof. Dr. Heinrich Nöth
President of the Bavarian Academy of Sciences
Distribution of turbulent energy in a mixer
Simulation of a lipid membrane
Charge density in a gallium nitrate crystal
Turbulent flow over an airfoil
Four speed profiles on an airplane
Flow profiles on an airfoil
Minister President Dr. Edmund Stoiber in conversation with Prof. Dr. Heinz-Gerd Hegering in the computer room at the LRZ