Test offers in and around GSI's Green IT Cube

Computing power meets sustainability – high-performance computing in the Green IT Cube

When data becomes a crucial resource, computing power is needed, and often a lot of it. This is exactly what is possible at GSI’s Green IT Cube in Darmstadt, where high-performance computing (HPC) meets sustainable technology.

What was previously used primarily in cutting-edge research is now becoming increasingly relevant for small and medium-sized enterprises: for example, in the development of new materials, in process optimisation or for data-intensive product innovations.

 

In this interview, Mohammad Al-Turany, Group Manager of IT Scientific Services and Research at GSI, and Helmut Kreiser, Head of Green IT Cube at GSI, explain what lies behind the Green IT Cube, how HPC works in concrete terms, and why powerful, energy-efficient computing is key to digital competitiveness today.

What exactly is high-performance computing, and why is it relevant for smaller businesses?

Al-Turany: High-performance computing (HPC) refers to the use of powerful computing systems to process complex calculations and data analyses that cannot be performed efficiently with conventional computers.

HPC is relevant for smaller companies (SMEs) because it gives them access to resources for demanding applications such as simulations and/or data analysis, areas that are often unaffordable due to high infrastructure costs.

How does an HPC consultation work in practice – from the initial discussion to implementation?

Al-Turany: At the beginning of an HPC consultation, the existing workflow is analysed to identify computing, storage and I/O requirements. Benchmarking and scaling tests are then carried out to identify suitable architectures, parallelisation strategies (e.g. MPI, GPU, containerisation) and resource models. Based on this, an optimised HPC concept with a suitable scheduler configuration (e.g. Slurm) and data management strategies is developed and implemented. Finally, integration, performance optimisation and user training are carried out to ensure efficient operation and monitoring of the systems.

What challenges do you encounter at SMEs when it comes to data-intensive applications or simulations?

Al-Turany: Many small and medium-sized businesses face very practical technical challenges when it comes to data-intensive applications or simulations. In most cases, they lack the necessary infrastructure, such as their own high-performance computers or sufficiently large storage systems. However, modern simulations or AI models often require parallel processing, fast I/O environments and large data volumes. Most SMEs are also unfamiliar with Slurm, containers, distributed file systems or batch systems. A lack of expertise in parallelisation – for example with MPI, OpenMP or GPU computing – makes it even more difficult to get started. In addition, there is often a lack of suitable backup or archiving solutions for securely transferring, storing and evaluating data. Data protection poses additional requirements, especially when it comes to customer data.

In addition to the technical challenges, there are also organisational hurdles. SMEs rarely have IT specialists with HPC or data science experience. The operation of computing clusters or large data pipelines requires continuous administration and monitoring, which is not covered in many companies.

The Green IT Cube is considered a particularly energy-efficient computing infrastructure. Can you explain what makes it stand out in technical and ecological terms?

Kreiser: It is characterised by an overall concept comprising the structural arrangement of all components, a compact building and innovative cooling technology: passively cooled rack rear doors use the server fans to directly cool the hot exhaust air, evaporative cooling towers are used to generate cooling water, and the absence of large air circulation fans reduces energy consumption. In addition, power is supplied via two independent feeds from the German high-voltage and extra-high-voltage grids – without UPS or emergency power systems.

A steel structure in high-bay warehouse design was chosen for the server core across six levels. This enables resource-saving energy use during operation (PUE ~ 1.07), reduces investment costs and ensures a high degree of recyclability of materials at the end of their life (grey energy).

What developments do you expect to see in the field of sustainable high-performance computing in the coming years – and how are you preparing for them?

Kreiser: The use of waste heat from data centres will become increasingly important. In combination with heat pumps, it can be used to heat low-energy buildings, for example. The low and medium temperature ranges achieved in this way can also be used in other areas. The use of directly liquid-cooled server systems (DLC) is becoming more widespread for high-performance systems in data centre environments. This results in higher waste heat temperatures (50–60 °C), which can be fed into district heating networks via heat pumps, for example. This enables the integration of sustainable data centres into urban infrastructures. Since it was commissioned in 2016, the Green IT Cube has been heating an office and canteen building using a heat pump and building core activation.