Supercomputers aid research at German institutes

The SX-ACE is well suited for scientific computing that requires super high speed parallel processing and advanced simulations using large scale data. The University of Kiel is using the product for simulations of climate change and oceanic phenomena. At Alfred Wegener Institute (AWI), a member of the Helmholtz Association of German Research Centers, the SX-ACE is being used for the numerical modelling of oceanographic physics and chemistry, as well as climatic variations particularly in the Earth’s polar regions.

The High Performance Computing Center Stuttgart (HLRS) is capitalising on the system for research, product development, and a spectrum of industrial applications in automotive, material design, and chemical fields through academia-industry collaborations.

The SX-ACE is a new vector supercomputer equipped with a multi-core vector CPU, which enables the world’s top level single-core performance of 64 GFLOPS and the largest memory bandwidth per core of 64 GB/s. Its performance per rack has improved ten times over the previous model, with a rack computing performance of 16 teraFLOPS (hereinafter ‘TFLOPS’) and a memory bandwidth of 16 Tbytes/second.

It is especially suited for scientific and engineering computing applications and data intensive applications that need high speed processing of big data. It achieves high sustained performance in various simulations - for weather forecasting, analysis of global environmental changes, fluid dynamics analysis, nanotechnology, development of new materials and others. Using NEC’s LSI technology, a high density packaging design, and high efficiency cooling technology, the SX-ACE also reduces power consumption by 90% and requires 20% of the floor space of the existing model.

University of Kiel

The University of Kiel maintains a close partnership with GEOMAR Helmholtz Centre for Ocean Research Kiel and offers high performance computing environments for cutting edge scientific research. GEOMAR conducts interdisciplinary investigations of all aspects of marine sciences, ranging from ocean circulation and climate dynamics, marine biogeochemistry, and marine ecology to dynamics of the ocean floor.

The SX-ACE with 256 nodes (maximum theoretical performance of 65.5 TFLOPS) introduced by the University of Kiel at this time is expected to significantly contribute to the numerical modelling of changing ocean and atmosphere climate due to global warming.

“It is now possible to undertake high resolution modelling of the Earth's systems involving complex physics and biogeochemistry without significant code tuning efforts. In particular, as compared to the previous system, we experienced a reduced time to solution with a higher effective performance. Furthermore, the required power consumption is lower than expected, facilitating the advancement of ocean and atmosphere sciences,” said Dr Andreas Lehmann, GEOMAR.

“Likewise, complex quantum chemical simulations based on the commercial software package Gaussian or paleoclimate simulations in geosciences over thousands of years, could be continued almost without any code modifications, and the new system already attracts new users with applications e.g. from theoretical physics and astrophysics. The energy consumption per compute operation of the new SX-ACE has been reduced by a factor of about 20 compared to its predecessor NEC SX-9. This is an essential step forward as computing centres are very concerned about total operating expenses, of which electricity and cooling have a significant share,” said Dr Holger Marten, Head of the University Computing Center.

Alfred Wegener Institute

Alfred Wegener Institute, a member of the Helmholtz Association of German Research Centers, focuses on research issues mainly related to Arctic and Antarctic regions, including oceanic physics, chemistry and ecology, as well as climate change. The SX-ACE with 32 nodes (maximum theoretical performance of 8.2 TFLOPS) introduced by AWI at this time is planned to be used mainly for the numerical modelling of the paleo climate using coupled earth system models, as well as the formation/decline of sea ice, and their interactions with climate change on a global or regional scale.

“AWI scientists have been successfully working with vector processors for several generations of high performance computers. Many models have been optimised for vector cores, and for some scientific studies (e.g. paleo climate) these processors are better suited than massively parallel set-ups. Therefore, AWI decided to install the SX-ACE in addition to a MPP cluster when replacing its outdated SX-8R. The 32 nodes of the new SX-ACE provide twice the computational power (in total and per-core) compared to 14 SX-8R nodes while using less than 20% of the electrical power. A 500 TB ScaTeFS WORK file system increases the available disk space by more than a factor five. Researchers were able to continue their work with little effort as no porting of model codes was necessary,” commented Dr Dirk Barbi of AWI.

HLRS

HLRS offers supercomputing resources for private companies in various fields in addition to academic institutions. Its computing environment is designed for scientific and engineering tasks through academia-industry interactions, making the centre one of the leading European facilities that encourage applications of high performance computing in engineering. The SX-ACE with 64 nodes (maximum theoretical performance of 16.4 TFLOPS) introduced by HLRS at this time is expected to be used for cutting edge industrial applications, such as fluid dynamics simulations in automotive design, computational physics modelling needed in designing novel materials, and computational chemistry at the core of drug discovery and bioinformatics.

“With the upcoming new paradigm of vectorising hardware on standard computing platforms, increasing performance and reducing energy consumption, HLRS recognises the importance of the continuation of the vector platforms of NEC for future efficient computing in simulation and data analysis. Essential for technical computing is the high sustainable memory bandwidth of the system. The machine will help us to improve the implementation of various complex codes, and I am looking forward to future efficient high performance machines,” said Dr. Uwe Kuster, HLRS.

“Using our vector technology, NEC will pursue the development of the next generation of high-performance servers targeting industrial application fields and large-scale data analyses, in addition to conventional supercomputer fields, thereby leading the latest developments into the future,” said Tomoyasu Nishimura, General Manager, IT Platform Division, NEC Corporation.