High-performance computing: Power from the fastest computer in Germany

In this article you will read about,

• why every company can now use the fastest computer in Germany,
• how cars can be developed faster with high-performance computing and,
• how to save a lot of money with HPC as a Service.

It is equipped with 720,896 CPU cores and has been on the campus of the University of Stuttgart since early 2020: Hawk – the fastest computer in Germany. According to its operator, the super computer performs 27 quadrillion computing operations per second and has a main memory capacity of 1.4 petabytes. That means it computes around 40,000 times faster than a classic PC and stores 250,000 times more information. With these performance figures, the digital hawk is not only top of the class in Germany, but also ranks among the Top 10 most powerful HPC installations worldwide.

But what exactly is meant by the term high-performance computing? "Basically everything that individual computers or servers cannot compute or can only compute very slowly," says Fabian Placht, HPC expert at the Open Telekom Cloud. The reason for the significantly higher speed of the supercomputers: "While conventional computers process their workloads in sequence, HPC clusters or supercomputers process them in parallel.” In this way, high-performance computers can calculate complicated models within minutes or hours, which would take a desktop PC several weeks or even months to do.

Users utilize HPC primarily to process huge amounts of data in a very short time, to simulate virtual variants of a certain basic constellation, and to set up or test prototypes. In this way, new products can not only be developed faster, but also be made more valuable and secure.

For example, in the automotive industry: Using CAE (computer-aided engineering) software, developers simulate how newly designed models react to air flows. In a virtual wind tunnel, they calculate how the vehicle behaves at 120 km/h. And they determine where the air is swirling, which design elements slow down the vehicle, and which improve its aerodynamics. The necessary calculations are not performed directly in the CAE software, but in the background on a high-performance computer.

Other industries also rely on high-performance computing: In the pharmaceutical industry, for example, medicines can be developed more efficiently. Product developers can simulate the effects of the composition of various active ingredients on the human organism much faster than with traditional computing power. Or which interactions can occur with existing medicines. In this way, new medicines or vaccines can be tested more easily by test subjects and new diseases can be treated more quickly.

In medical research, the supercomputers also perform useful services and help save lives – for example, by improving and accelerating imaging diagnostics. Using HPC resources, extremely large quantities of storage-intensive MRT/CT images of human organs can be compared and analyzed within a very short time. Based on the results, doctors can then detect diseases such as Alzheimer's or cancer earlier and treat patients more effectively.

However, the high performance is expensive. The state and federal government have spent 38 million euros on the Hawk, which is also available to industrial users via the public-private partnership, Höchstleistungsrechner für Wissenschaft und Wirtschaft (HWW) GmbH. The HWW is a joint venture between the state of Baden-Württemberg, the Karlsruhe Institute of Technology (KIT), the University of Stuttgart, Porsche, and T-Systems. Their goal: In addition to science, industry in particular should benefit from the high-end computer. Start-ups and SMEs, but also large corporations can use the computing capacities. For example, with an offer from the Open Telekom Cloud: In the form of HPC as a Service, customers can access the computing power of the more than 700,000 processor cores of the Hawk from the public cloud and model or simulate as required.

If companies need less computing power, they can also go one size smaller: From one to 1,000 processor cores per computing run, the Open Telekom Cloud offers for high-performance computing in the form of bare metal servers – bare metal hardware equipped with powerful CPUs. The servers can be interconnected as clusters, with the standard InfiniBand protocol ensuring fast data exchange. Via a hardware interface, it allows low latencies over short distances and thus ensures fast results.

In addition to CPUs, companies can also use powerful graphics processors (GPUs) for high-performance computing in the Open Telekom Cloud, for example to train artificial intelligence more effectively or to evaluate large data analyses in real time. The choice of the most suitable processor type depends on how data-driven the applications used are or on which software code they are based.

Ultimately, HPC as a Service offers the same advantages as computing in the public cloud. Scalable resources can be used as needed within seconds. Billing is based on the pay-as-you-use model – customers only pay for the services they actually use. In addition, the latest technology is always available. The developments in supercomputing are extremely fast moving. High-performance computers become obsolete within two to three years. Therefore, it is hardly worthwhile for companies to invest in these technologies – the acquisition costs are too high, the depreciation periods too short.