The company JUWELS develops new generation of flexible, modular supercomputers. The modular approach means that the computing systems can be easily adapted, and they can be used for a range of tasks. The first module is the fastest German developed computer on the TOP500 list of the fastest computers in the world. The TOP500 project ranks and details the 500 most powerful non-distributed computer systems in the world. The project was started in 1993 and publishes an updated list of the supercomputers twice a year.
A driver for super-computers is digital transformation projects within companies and the need to undertake various big data applications and computational simulations. The need for more sophisticated analyses is driving the need for more powerful computers, as is the case with the JUWELS model.
According to Thomas Lippert, Director of the Jülich Supercomputing Centre, and reported to EE News Europe, the use of a modular approach is important in order to deliver affordable and energy-efficient technology for businesses.
Lippert is championing the concept of an adaptable design, which he calls “Smart Exascale”. This is in keeping with the European Union research project DEEP. Since 2011, experts from 16 European partners have been working on DEEP (Dynamical Exascale Entry Platform – Extreme Scale Technologies) projects funded by the European Union.
The module is energy-efficient through the use of hot water cooling. This type of cooling is able to cool the majority of the waste heat using hot water drawn from the outside air. The design does not require additional cooling generators.
The modular concept involves developing computerized systems that consist of several specialized modules which can be flexibly combined through appropriate operating software. As an example, French IT company Atos has developed a cluster model which runs with software developed by the German company ParTec. The computer uses Intel Xeon 24-core Skylake CPUs. This system provides a maximum performance of 12 petaflops (flops are “floating point operations per second”, in this case the performance power is best expressed as the computing power of 60,000 PCs).
