Technical Innovations and New Services

New high-performance computers, practical storage solutions, and additional useful services for the digitization of research and education: Those relying on the IT services of the Leibniz Supercomputing Center can look forward to exciting developments in 2024.

The supplemental system of SuperMUC-NG, known as Phase 2 or SNG-2, has been installed in the computing cube at the Leibniz Supercomputing Center (LRZ). The high-performance computer (HPC) has been undergoing testing by researchers and early users since January 22, with LRZ specialists preparing it for operational use. With 960 Graphics Processing Units (GPUs) from Intel, the system accelerates traditional calculations and enables the implementation of Artificial Intelligence (AI) methods. To facilitate researchers in getting acquainted with the necessary workloads and tools offered by SNG-2 and to allow for immediate experimentation, the Computational X Support (CXS) mentorship team at LRZ has collaborated with data and AI experts from the Big Data & Artificial Intelligence team. Additionally, in cooperation with chip manufacturer Intel, workshops and training sessions on handling SNG-2 are being prepared for the first half of 2024.

However, the supercomputer is not the only innovation that researchers and students at LRZ can anticipate. In 2024, the agenda includes the acquisition of innovative technologies as well as the introduction of new services, some of which are developed in collaboration with Bavarian universities. An overview:

HPC Systems and Storage

In addition to the operational launch of SNG-2, the procurement of a new Linux cluster is currently underway, under the project name CoolMUC-4. The German Research Foundation (DFG) has approved the procurement request for this project. A significant portion of this system will incorporate GPUs, thereby enhancing scientific methodologies. Traditional calculations can be combined with pattern recognition and AI models, allowing simulations to be extended through surrogate models and emulators. Similar to SNG-2, CoolMUC-4 will gradually be equipped with tools for processing research data.

Also demonstrating increased efficiency and strength is terrabyte, the high-performance platform for the analysis of Earth observation data established and operated by LRZ in collaboration with the German Aerospace Center (DLR). It was equipped with additional computing and storage capacity in the fall, now achieving 3.4 trillion floating-point operations per second. A substantial amount of DLR data can be processed using pre-installed applications. With these "Analysis Ready Data," terrabyte supports research on climate change, urban development, and the utilization of forest, land, and water areas. Since AI relies on extensive datasets, storage becomes crucial, and rapid data transport is essential for energy efficiency. Therefore, the capacities of terrabyte are being expanded further. Additionally, LRZ specialists are planning to acquire additional servers to optimize the Data Science Storage (DSS) and the Data Science Archive (DSA). The Archive and Backup System (ABS) is currently under close scrutiny and is expected to be supplemented with more powerful systems in 2024.

Artificial Intelligence (AI) and Quantum Computing

Not only are the HPC resources now GPU-enabled for AI, but LRZ also operates its AI Systems to meet the diverse needs of AI in science. These systems facilitate computer vision and the training of AI models with data. Additionally, LRZ hosts the same system for the Munich Centre for Machine Learning (MCML). If necessary, the capacities of both systems can be increased through the other. Both systems will also be supplemented with additional processors in 2024.

A technical upgrade is also making the Cerebras CS-2 system more versatile for research applications. Its giant chip is designed for rapid data flow and data training. The LRZ Big Data & Artificial Intelligence (BDAI) team is currently evaluating new functionalities and accompanying tools before releasing the system again for researchers.

In 2023, the Quantum Integration Centre (QIC) at LRZ achieved a quantum system based on trapped ions, complementing existing systems using superconducting technology. What can be disclosed is that in 2024, additional quantum technology will be introduced, integrating it as an accelerator into LRZ's supercomputers and making it available to partner institutions within the Munich Quantum Valley (MQV). This includes expanding the Munich Quantum Software Stack. The Bavarian Quantum Portal is taking concrete shape for accessing quantum computers, aiming to allow European researchers access to quantum technologies. Until January 22, 2024, the EuroHPC Joint Undertaking tender was open, with LRZ vying for one of six European hybrid quantum computers based on superconductors. The undertaking is now evaluating concepts and planning the procurement process.

Security and Services

New technologies emerging from and for research at LRZ will, in the medium term, lead to new IT services and security measures. In 2024, LRZ will intensify collaboration with the Digitalverbund Bayern, working with the organized universities within it to develop strategies against cyberattacks and for IT security and more. It is already confirmed that BayernShare, the platform for file exchange, will be secured with Two-Factor Authentication (2FA). A web-based access with 2FA will also be implemented for both the existing and new Linux clusters.

The Center for Virtual Reality and Visualization (V2C) will introduce Video-on-Demand and streaming services in 2024. This will allow research institutions to more easily disseminate videos on the internet or make lectures and workshops accessible to a large audience—a milestone for science communication of research projects.

Through participation in European projects such as EXA4MIND and collaboration on the National Research Data Infrastructure (NFDI), LRZ is developing many tools for managing research data and workflows with large volumes of data. In the spring, the LRZ FAIR Data Portal will be made available for testing to selected partners. Researchers will find practical ways to make datasets from LRZ Storage searchable and publish them according to the widely accepted FAIR principles. If the system proves successful, it will be offered to a broader user base.

Additionally, the LRZ will modernize its website and increase visibility on social media. In addition to Linkedin and X (formerly Twitter), the PR team has established a presence on Mastodon and plans further videos for the LRZ Youtube channel. It's worth checking out: #discoverLRZ.

Energy and Environment

LRZ specialists, particularly those planning high-performance systems and computer resources, require space. Additionally, to enhance energy efficiency and sustainability, a sophisticated infrastructure for cooling with hot water is needed. LRZ is planning an expansion building where not only the power supply for future supercomputers but also innovative technologies like quantum systems will be accommodated.

The initiative becomes more specific with the goal of generating energy on-site. In 2024, the first photovoltaic modules will be installed on the roof of Institute Building 2, and the following year, the southern LRZ building will also be equipped with them. Concurrently, LRZ is preparing a management system to collect and analyze energy and environmental data during operations. This system will help reliably determine legally mandated metrics and provide information to optimize IT services and computer resources for research and education in terms of sustainability. (vs)