Computer technology of the future
Computer technology is changing rapidly. That is why the LRZ is researching innovative hardware and software for energy-efficient operations at the computing centre and to provide users with the best possible computer technology.
How will the computers of the future work? The activities of our Future Computing Team centre around this question. The LRZ’s specialists are working with scientists from other institutes and experts from tech firms to explore state-of-the-art components for use in high-performance computing (HPC). In addition to working methods and developing benchmarks for operations, the focus here falls on the energy requirements of processors and systems.
The LRZ has created a test environment called BEAST, the Bavarian Energy, Architecture, and Software Testbed, for sharing experiences on innovative hardware with researchers, for co-design projects with start-ups and companies, and for taking part in research projects related to supercomputing: In times of growing complexity in IT, we want to get a comprehensive understanding of technological concepts so that we can ensure that the LRZ’s computer resources are energy-efficient, practical, and useful in the daily application of science and research.
Integrated supercomputing
Dennard Scaling has already been exhausted, and Moore’s law is pushing its limits. It will no longer be possible to double the performance of processors in a short time, and the increasing energy requirements of chips and systems also need to be brought under control. The consequences of this are felt when planning supercomputers: Hardware is tailored to application fields for the sake of greater performance and economy. As a result, HPC systems today integrate a great deal more components than ever before. The LRZ’s Future Computing Group and scientists from the Technical University of Munich (TUM) state their views on the development of supercomputing in a position paper on the future of supercomputing and integrated HPC systems.
Exploring supercomputing
The LRZ has installed the Bavarian Energy, Architecture- and Software Testbed ( or BEAST for short) to comprehensively test processors, accelerators and other components: It collects performance figures from the latest hardware and creates benchmarks for evaluating components and software. This involves experimenting with clock frequencies and other control elements to sound out the energy requirements. And this is where the plans for the next supercomputers are developed or planned components are optimised. BEAST is not an LRZ service but is made available to selected research groups.
Improving use
The Future Computing Team works closely with Computational X Support (CXS) when working on BEAST: This is because planned components can be tested in the test environment and, if necessary, optimised or stabilised for use at LRZ. BEAST can also be used to adapt and improve software and scientific code: Experience that is also highly valued by in tech firms – the future of supercomputing lies in co-design and close technical coordination with potential user groups. This also increases user-friendliness.
Monitoring with DCDB
Thousands of sensors are installed inside the LRZ’s supercomputers and the computer cube, which record values from functions or the environment on a daily basis. They are collected in Data Centre Data Base (DCDB), an openly available, plugin-based monitoring tool from the LRZ that is now used as reference at many supercomputing centres. DCDB uses data to reveal the adjustments that can be made to optimise the technology, HPC and the computing environment. Tools such as DCDB also lay the foundations for developing automated control systems.
Familiarising students with technologies
Students at both Munich universities can familiarise themselves with various technology concepts as part of a BEAST internship. BEAST currently includes processors from AMD, Cerebras, Fujitsu, Intel, NVIDIA and more, as well as network and memory components. Students are given practical tasks over the course of a semester to compare or change their performance values.
-
Tools for exascale computing
The LRZ cooperated with universities and other computing centres as part of a European research project to develop the tools for modular supercomputing at exascale level. The result was a software package for controlling and using supercomputers and for increasing energy efficiency.
-
Economical calculations
REGALE was the name of a research project in which the LRZ, together with 15 universities, computing centres and companies, explored strategies for managing the energy consumed by supercomputers and also produced a practical toolbox for sustainable high-performance computing: This now contains freely available, tried and tested tools and software for running computing centres sustainably, as well as for sustainable coding and programming.
-
Examining accelerators
In addition to graphics and quantum processing units (GPU/QPU), there are other, specialised components that can accelerate supercomputing, such as the SmartNIC or SmartSwitches. These are being examined in ScalNext, and software solutions that allow data management and control measures to be transferred from the processors to the network of a HPC system are also being subjected to testing.
-
AI & Data ScienceBig Data & AI11.02.2025Kick-off for BayernKI
With BayernKI, LRZ and NHR@FAU offer fast and flexibly scalable access to high-performance infrastructure.
-
SupercomputingFuture Computing13.12.2024HPE builds next-generation supercomputer at the LRZ
New supercomputer “Blue Lion” is part of the German national HPC infrastructure of the Gauss Center for Supercomputing…
-
AI & Data ScienceBig Data & AI10.12.2024New Technology Speeds Up Fight Against Online Harassment
With the CS-2 system from Cerebras, AI models for the recognition of hate speech on social media platforms can be…
Curious about the future of computing?
If you would like to work with the LRZ team to advance or explore innovative technology, you should get in touch with our computer technology specialists.
