en.Wedoany.com Reported - The German Climate Computing Center (Deutsches Klimarechenzentrum, DKRZ) has deployed the VAST AI operating system to modernize its high-performance computing environment for climate research.
DKRZ is Germany's dedicated high-performance computing center focused on climate science, providing computing and data services for German and international climate research projects. The center manages over 250 petabytes of climate data, operates the World Climate Data Center, and plays a key role in the global climate research ecosystem through initiatives such as the Earth System Grid Federation.
As climate models become increasingly complex and artificial intelligence begins to reshape scientific workflows, research computing centers face pressure to support more users, more data, and more demanding development environments. For DKRZ, this challenge extends beyond the largest simulation outputs. Researchers also require reliable access to their daily work environments, including home directories, software stacks, Python environments, shared project spaces, code repositories, and visualization scripts—the foundations upon which scientific research is conducted.
DKRZ chose VAST to replace and modernize its high-performance computing home directory and software environment, aiming to provide researchers with a more reliable daily work foundation while offering the infrastructure team a simpler, more efficient management platform.
Michael Boettinger, Head of Visualization and Outreach at DKRZ, stated that climate science relies not only on raw computing power; researchers need reliable access to environments for developing models, building software, preparing workflows, and analyzing results. With VAST, DKRZ can provide a modern home directory and software environment, offering users greater capacity, better reliability, and fewer technical limitations. Keeping the primary environment continuously available is a significant improvement.
For DKRZ, the decision to deploy VAST was driven by its complete platform capabilities, including snapshots, data reduction, monitoring, ease of management, non-disruptive upgrades, and effective support for small-file-intensive workflows. These features are particularly critical in research environments, where users often create numerous similar but not identical software and Python environments, generating a large number of small files that are inefficient to manage in traditional systems.
Anna Fuchs, HPC I/O and Storage Engineer at DKRZ, noted that they were not only pursuing performance but also needed a future-proof system with the right combination of features, reliability, and management simplicity. Snapshots, monitoring, compression, and data reduction are all important, but the value of VAST integrating these functions into a single platform lies in enabling users to focus on their work rather than on technical limitations and potential downtime.
DKRZ has already observed significant effects from VAST's data reduction capabilities. In the production user environment, the center has observed an average reduction factor of approximately 5.5x, with specific software development and build environments achieving reduction factors of around 20x or higher. This allows DKRZ to provide more usable capacity for researchers' workflows without simply adding hardware.
The platform also enables DKRZ to introduce shared environments between individual home directories and the global software stack, allowing project teams to maintain common software and Python environments without duplicating similar files across users.
VAST AI OS enables DKRZ to provide researchers with uninterrupted access to critical research environments. Through non-disruptive upgrades and resilient system design, VAST, together with VAST DataSpace, ensures that home directories, scripts, and software environments remain continuously available. Its high IOPS architecture effectively handles small-file-intensive workloads such as Python environments and build systems. Built-in compression, deduplication, and similarity-based reduction increase effective capacity. Rolling software updates allow maintenance and upgrades without routinely taking user environments offline. Integrated monitoring, snapshots, and cataloging functions provide administrators with greater visibility and control.
DKRZ is also exploring other use cases for VAST, including internal services and code-related workflows that require reliable access for researchers and collaborators. The center is evaluating additional platform features to consider how VAST can support future research infrastructure needs.
Christopher Huggins, Managing Director of AI and HPC for EMEA at VAST Data, stated that research computing centers are being asked to support more users, more data, more complex software environments, and increasingly AI-driven workflows, often with limited teams and budgets. DKRZ understands that modern research infrastructure relies not only on computing; the data layer is not just a backend storage issue but the foundation upon which researchers develop, collaborate, and transform computation into scientific insight.
As climate science enters a new era of higher-resolution modeling, larger datasets, and AI-assisted research workflows, DKRZ is leveraging VAST to build a more reliable, efficient, and flexible foundation for scientific computing, with plans to scale further.
VAST Data is an AI operating system company that powers the next generation of intelligent systems through a unified software infrastructure stack built to unleash the full potential of AI. VAST AI OS integrates foundational data and computing services with agent execution into a scalable platform, enabling organizations to deploy AI agents globally, facilitate their communication, reason over real-time data, and automate complex workflows. VAST is built on its breakthrough DASE architecture, the world's first true parallel distributed system architecture, designed to eliminate trade-offs between performance, scale, simplicity, and resilience.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
