The power of two hearts: HPC and supercomputing as required

In this article you will read about, 

• how Telekom combines cloud resources with supercomputing capacities in Stuttgart,
• how companies can benefit from the combination of high-performance computing from the public cloud and supercomputing
• and what makes this combination currently unique on the market.

Would you like a little bit more? There are workloads that are too large even for high-performance computing (HPC) in the scalable public cloud. For example, when very high speed is required for highly complex calculations or analyses. Then the capacities of a highly specialized, classic supercomputer can help. But what is the difference between HPC from the public cloud and a supercomputer?

"In classic cloud environments such as the Open Telekom Cloud, the technology currently in use reaches its natural limits with around 1,000 x 86 cores simultaneously assigned to a problem," says Alfred Geiger, Managing Director of Höchstleistungsrechner für Wissenschaft und Wirtschaft GmbH (HWW), the joint operating company of the High-Performance Computing Center Stuttgart (HLRS), T-Systems, Porsche AG and the Karlsruhe Institute of Technology (KIT).

One reason for the limit is the network: the computing cores must be able to communicate with each other at all times without blocking, with low latency and sufficient bandwidth, so that the cluster can work efficiently. The required packing density and communication infrastructure would go far beyond the scope of a general-purpose cloud environment. Geiger: "At the High-Performance Computing Center in Stuttgart, we can currently process workloads with up to 180,000 cores, i.e. up to 180 times more complex or faster. This is due to the architecture of the hardware, which is completely designed for supercomputing."

This makes the Stuttgart computing capacities ideally suited for very complex problems that have to be solved in a very short time. The HLRS resources offer an important unique selling point: Companies can book them - just like the high-performance computing resources from the Open Telekom Cloud - according to their needs. "This gives us a flexible supercomputer that scales with demand," says Alfred Geiger. "Companies only pay for the supercomputing resources they need. This is unique in the market to date. This is because other providers generally provide supercomputing resources only on a dedicated basis. This can sometimes result in significantly higher costs."

The High-Performance Computing Center in Stuttgart is directly connected to the Open Telekom Cloud data centers in Magdeburg and Biere via a dedicated network connection. This enables companies to easily combine the resources of the public cloud and the supercomputing data center. 

Telekom thus offers the exact computing capacity that companies need for every conceivable complex problem: Applications with implicit codes, for example, which require very high I/O speeds, can be best implemented with HPC flavors in combination with fast nonvolatile memory express (NVMe) storage from the Open Telekom Cloud. 

Complex CAE applications from the Product Lifecycle Management (PLM) area, on the other hand, function optimally with the help of supercomputing capacities from Stuttgart. "The clever combination of both – HPC and supercomputing – offers many advantages," says Max Guhl, Chapter Lead – HPC on Open Telekom Cloud. "Companies can order highly demanding workloads within a data center network. And they always get the right solution for every application, which they can use as needed."

Stuttgart's HLRS was founded in 1996 as the first national high-performance computing center in Germany. Its capacities are used by more than 1,000 German academic institutions and countless companies, mostly for research and development purposes. The organizational framework for industrial and commercial use is provided by Höchstleistungsrechner für Wissenschaft und Wirtschaft GmbH (HWW), in which T-Systems, among others, has a stake.