High-performance computing center Stuttgart

from Wikipedia, the free encyclopedia
Logo of the high-performance computing center Stuttgart (HLRS)
Building on the Stuttgart-Vaihingen research campus

The High Performance Computing Center Stuttgart ( HLRS ) is a computing center that offers science and industry access to supercomputers . It was founded in 1995 under the umbrella of the Computer Center of the University of Stuttgart (RUS) and has been the first German federal high-performance computer center since 1996. The HLRS has been an independent central institution of the University of Stuttgart since 2003. It houses the fastest supercomputer in Europe (as of 02/2020), an HPE Apollo 9000 system called Hawk and a maximum computing power of 26 Peta - FLOPS . It went into operation in 2020 and the cost was around 38 million euros.

The HLRS operates computers, looks after German and European users and offers an intensive training program for its users. In addition, the HLRS conducts its own research in the field of high-performance computing. The focus is on the topics of scalability, performance optimization, big data, green IT, in the application areas of health, environment, energy and mobility, as well as the investigation of social, political and philosophical aspects of simulation. The HLRS has received several international awards for its research. The HLRS has been making its computers and services available to industry since 1995 through a public-private partnership with Porsche and T-Systems. With the founding of the Automotive Simulation Center Stuttgart (ASCS), the HLRS has created a development and research platform with the automotive industry, on which pre-competitive research in close coordination between automotive manufacturers, suppliers, computer manufacturers and software companies contributes to the optimal use in improving mobility.

Supercomputers

Cray XC40 "Hazel Hen".
Cray XE6
NEC SX-9
View into the data center
NEC SX-ACE (left) and NEC Cluster (right, white).

Since October 2015, the core system of the HLRS has been a Cray XC40 System Hazel Hen with a maximum output of 7.42 Peta-FLOPs. Since February 2020 Hazel Hen has been replaced by the HPE Apollo 9000 System Hawk with around 26 Peta-FLOPs (Peak Performance). The Hawk supercomputer is (as of February 2020) the fastest supercomputer in the EU and fifth worldwide .

Other systems are the NEC cluster (Vulcan and Vulcan2). In addition, a large number of smaller systems are operated that are used for test purposes (NEC SX-ACE) or e.g. B. can be used for development and visualization. A cave with five projection surfaces is available for the three-dimensional display of simulation results.

history

High-performance computing in Stuttgart goes back to a long tradition, which has its origins in the fields of aerospace ( John Argyris ) and nuclear technology ( Roland Rühle ). The integration of these activities in a central facility goes back to the acquisition of the most powerful supercomputer Cray-2 , a vector computer , by Lothar Späth in 1986.

When the HLRS was founded in 1996, the HLRS had two different systems with an NEC SX-4, also a vector computer, with a peak performance of 64 gigaflops and a Cray T3E / 512 with 461 gigaflops of theoretical computing power, which were equally among the best worldwide. The first director of the HLRS was Roland Rühle, who established the HLRS as part of the computer center of the University of Stuttgart (RUS). The responsible division manager - and thus the first head of the HLRS - was Alfred Geiger. In 1999 the HLRS was recognized by the American National Science Foundation (NSF) for its work in the field of high-performance distributed computing. In 2000, an NEC SX-5 / 32M2 and a Hitachi SR-8000 were procured, each with a theoretical computing power of 128 gigaflops.

In 2002, after Roland Rühle's retirement, the HLRS was separated from the computer center of the University of Stuttgart as an independent facility. Michael M. Resch took over the management. In 2003 the HLRS won the HPC Challenge in the USA. The NEC SX-6 (six nodes, 440 GigaFlops Peak Performance) purchased in 2003 was again a vector computer. In the same year, an AMD Opteron cluster and an Intel Xeon cluster were also procured. The installation of the vector computer NEC SX-8 (12.8 TeraFlops Peak) took place in 2005, for which a new building was erected. In 2007 the HLRS was a founding member of the German Gauss Center for Supercomputing (GCS), in which the three German federal high-performance computing centers are united. In 2010 a new energy supply building was built. In 2011, the Cray XE6, the first computer with a performance of more than one Peta-FLOP, was installed, which was replaced in December 2014 by the Cray XC40 "Hornet" system (peak performance: 3.8 Peta-FLOPs). Since 2012, the HLRS has also provided European users with computing time as part of PRACE. In October 2012 the HLRS put a new research building into operation, in which a new 5-sided CAVE was also installed.

In organizational terms, the HLRS was a member of a network of three centers within the University of Stuttgart from 2012 to 2016, which were combined in the Information and Communication Center of the University of Stuttgart ( IZUS ).

The new Hawk supercomputer was inaugurated on February 19, 2020 .

Access and operation

The mainframes are operated by the HLRS and are accessible to different users.

Science: The allocation of computing time is carried out by a scientific steering committee (LA). Half of its members are nominated by the German Research Foundation (DFG) and half by the Baden-Württemberg State Rectors' Conference (LRK). Applications are submitted to the HLRS and are forwarded from there to the LA. As part of the all-German cooperation in the Gauss Center for Supercomputing (GCS), very large applications are assessed by an all-German committee made up of the committees of the three member centers of the GCS. Outstanding projects can be accessed via the European HPC partnership PRACE. In this way, too, an assessment is carried out by a scientific committee.

Industry: The HLRS computers are also rented to industry via the Höchstleistungsrechner für Wissenschaft und Wirtschaft GmbH (hww). In addition to HLRS, which operates the computers, as well as KIT and the state of Baden-Württemberg, T-Systems , T-Systems Solutions for Research GmbH and Porsche are other shareholders. Pure production operations are offered and sold via the hww.

Commercial cooperation / test: For commercial use within the framework of a cooperation or for projects in which companies want to test access to the computer, access can take place within the framework of a cooperation. In this case, the cost of industrial use must be borne.

The use of the computers for military purposes is fundamentally excluded. The use of computers for matters that are affected by the German Foreign Trade Act and the corresponding embargo provisions (nuclear technology, rocket technology, ... for boycotted countries) is also excluded.

research

The HLRS conducts its own research in the field of high-performance computing. It was recognized for its research in the United States in 1999 and 2003. In around 35 projects, around 60 scientists work on problems from the field of simulation and high-performance computing. The main research topics of the HLRS are

  • power
  1. Performance analysis and optimization of the performance of computer systems
  2. Performance analysis and optimization of application program performance
  3. Performance analysis and optimization of data storage systems and file systems
  • Big data
  1. Handling of very large amounts of data
  2. Data analytics
  • Green IT
  1. Analysis and optimization of the energy consumption of computer systems
  2. Analysis and optimization of the energy consumption of applications
  • Parallel and distributed systems
  1. Cloud / grid concepts and applications
  2. Up-scaling algorithms and methods
  3. Parallel programming models and languages
  • Applications (together with application partners)
  1. Health: bone simulations and simulations of blood flow
  2. Environment: simulation of local effects of environmental influences
  3. Energy: simulation of energetic systems
  4. Mobility: Simulation of vehicle technology issues
  • Visualization
  1. Virtual reality
  2. Augmented reality
  • Society and politics
  1. High performance computing in school lessons
  2. Simulation in policy making
  3. Philosophical and sociological aspects of high-performance computing

The HLRS is currently involved in the excellence cluster "Simulation Technology" as part of the federal government's excellence initiative. The director of the HLRS is a Principal Investigator in the Cluster of Excellence. In addition, the HLRS is involved as a project partner in the Collaborative Research Center 716 “Dynamic simulation of systems with large numbers of particles”. In Europe, HLRS is a leading partner in a number of projects to develop new hardware and software technologies for clouds and supercomputers.

further activities

  • Education: The HLRS is involved in lecturing at the University of Stuttgart. The HLRS is integrated into the mechanical engineering courses through lectures through the director of the center.
  • Further education: The HLRS offers various courses on the subject of high-performance computing. Are treated z. B. Programming languages ​​from the field of high-performance computing ( C , Fortran ), parallel programming with MPI and OpenMP or efficient methods for solving systems of equations.
  • Cray Development Center: The HLRS has been working closely with Cray on the development of software since 2012.
  • Teraflop Workbench: The Teraflop Workbench is a cooperation between HLRS and the NEC company. The aim of the project is to enable existing applications with real computing power in the teraflops range.
  • Gauss Center for Supercomputing : The Gauss Center for Supercomputing is an amalgamation of the three federal high-performance computing centers in Jülich (Jülich Supercomputing Center / JSC), Munich-Garching (Leibniz-Rechenzentrum / LRZ) and Stuttgart (HLRS)
  • HPC Europe: The cooperation of different data centers in Europe promotes the exchange of scientists and enables the use of supercomputers in other countries.
  • PRACE - Partnership for Advanced Computing in Europe : prototypes and software for future petaflop systems

Awards

  • 1999 High Performance Computing Award from the US NSF: for the first coupling of two mainframes between Europe and the USA
  • 2003 HPC Challenge of Supercomputing
  • 2007 Microsoft Award "Early Contributor to Microsoft's Windows Compute Cluster Server 2003"
  • 2007 Selected Location, Germany - Land of Ideas
  • 2007 Invited Plenary Talk at SC'07 Reno / USA by the director of HLRS Michael Resch
  • 2009 Honorary doctorate from Donetsk National Technical University, Ukraine for the director of HLRS Michael Resch
  • 2010 ITEA Gold Award for the ParMa project
  • 2011 honorary doctorate from the Russian Academy of Sciences for the director of HLRS Michael Resch
  • 2012 The day after tomorrow maker prize from the state of Baden-Württemberg for the director of HLRS Michael Resch
  • 2013 2nd place in the prize of the Bosch Foundation "School meets Science" for the HLRS project "Simulated Worlds"
  • 2013 HPCwire 2013 readers' Choice Award for Best use of HPC Applications in Manufacturing for HLRS
  • 2014 honorary professor title of the Russian Academy of Sciences for the director of HLRS Michael Resch
  • 2014 Award of the HLRS project "Simulated Worlds" as "Particularly Prize-worthy" by the Education Foundation of the Kreissparkasse for the Esslingen district

literature

The research results obtained on the HLRS computers from a wide variety of specialist areas are published annually as a book in the series High Performance Computing in Science and Engineering by Springer-Verlag .

The work within the framework of the Teraflop Workbench and the Workshop on Sustained Simulation Performance is reported in the book series High-Performance Computing on Vector Systems (2005-2011) and Sustained Simulation Performance (since 2012), which are also published by Springer-Verlag is relocated.

The Tools for High Performance Computing series has also been published by Springer Verlag since 2008 through the workshop on Tools in High Performance Computing carried out jointly with the ZIH of the TU Dresden .

The HLRS publishes the Gauss Center for Supercomputing's magazine called Inside (Innovative Supercomputing in Germany) every six months , which is also offered as an online edition.

Cross-reference to article on the HPC-Europe funding program added

See also

Individual evidence

  1. ↑ http://stuttgart-journal.de/tp3/2018/11/15/hawk-der-schnellste-supercomputer-deutschlands/
  2. ↑ Press release Uni Stuttgart http://www.uni-stuttgart.de/hkom/presseservice/pressemitteilungen/2014/090_HLRS_Hornet.html?__locale=de
  3. ↑ Press release Uni Stuttgart http://www.uni-stuttgart.de/hkom/presseservice/pressemitteilungen/2015/79_hlrs_richtfest
  4. ↑ http://www.hlrs.de/systems/hpe-apollo-9000-hawk/
  5. ↑ http://www.hlrs.de/systems/cray-xc40-hazel-hen/
  6. ↑ http://www.hlrs.de/systems/hpe-apollo-9000-hawk/
  7. ↑ a b https://www.uni-stuttgart.de/en/university/news/press-release/Hawk-Supercomputer-Inaugurated/
  8. ↑ http://www.hlrs.de/systems/platforms/
  9. ↑ http://www.teraflop-workbench.de
  10. ↑ HPC Europe (website)
  11. ↑ http://www.prace-project.eu/
  12. ↑ Press release from the University of Stuttgart
  13. ↑ https://land-der-ideen.de/
  14. ↑ http://inside.hlrs.de

Web links

Commons : Höchstleistungsrechenzentrum Stuttgart  - collection of images, videos and audio files

Coordinates: 48 ° 44 '22.3 "  N , 9 ° 5' 49.9"  E