Vast Data, an AI operating system company, announced that the Swiss National Supercomputing Centre (CSCS), part of ETH Zurich, presented two technical papers at the 2026 Cray User Group (CUG) conference in Nice, France, highlighting Vast Data’s role in advancing trusted, high-performance infrastructure for modern AI and HPC workloads.The research shows how one of Europe’s leading supercomputing centres is helping define the next phase of AI and HPC infrastructure. As scientific computing becomes more data-intensive and AI moves into sensitive domains such as healthcare, national research centres need infrastructure that can protect regulated data, support secure shared access, serve new I/O patterns and deliver the performance expected from supercomputing-scale systems.

CSCS presented research on “Architectural Isolation for Sensitive Workloads: Enabling Trusted Research Environments on HPE Cray EX Systems,” which received the CUG 2026 Best Paper Award. The recognition highlights the importance of trusted research infrastructure within the supercomputing community, as HPC centres are increasingly asked to support sensitive data, regulated workloads and AI-driven research at scale. CSCS also presented “Evaluating the Operational Viability of Vast Data as a Scratch File System for Alps,” which examined the use of Vast infrastructure for HPC and AI workloads on Alps, CSCS’s HPE Cray EX supercomputer.

Advancing Trusted Research Environments for Sensitive AI Workloads
Trusted Research Environments are highly secure data and computing spaces that allow approved researchers to access and analyze sensitive data without that data leaving the controlled environment. As AI increases demand for larger-scale compute, CSCS is developing an architecture that enables sensitive workloads to run on supercomputing infrastructure while preserving security, compliance and near-native performance.

The CSCS TRE architecture is designed around three isolation pillars: network fencing through HPE Slingshot, storage encryption capabilities through the Vast Data Platform, and a circuit breaker mechanism. The work includes a PoC with Centre Hospitalier Universitaire Vaudois (CHUV) and the CHORUS platform that CHUV uses for secure data analysis and access, focused on training an EEG foundation model to support research into epilepsy surgery, SUDEP risk prediction and brain signal decoding.

“CSCS is committed to enabling research that delivers meaningful scientific and societal impact,” said Maxime Martinasso, associate director, head, engineering, CSCS. “As AI workloads become more important across healthcare and other sensitive domains, trusted infrastructure must evolve to give researchers access to the scale they need while maintaining the security and governance these environments require. Our work at CUG reflects an important step toward that goal.”

Validating Vast Data as Scratch Storage for HPC and AI
In a second CUG paper, CSCS evaluated Vast Data as scratch storage for Alps, testing whether the platform could meet the combined enterprise and HPC requirements of modern supercomputing environments, including encryption, multi-tenancy, QoS, network isolation, HPE Slingshot compatibility and performance for AI and HPC workloads.

CSCS selected a Vast system to evaluate whether the high-throughput storage layer used by active jobs could be delivered alongside the enterprise capabilities required for sensitive, shared research environments. In acceptance testing, Vast met CSCS’s required performance targets with multi-tenancy and encryption enabled, using native multi-protocol access through standard protocols, direct access to the HPE Cray Slingshot fabric, and no gateway layer. Application-oriented testing also showed Vast performing strongly vs. Lustre-based environments, with CSCS concluding that “Vast Data can be used as encrypted scratch for HPC/AI.”

“Vast was built for where infrastructure is going,” said Christopher Huggins, MD, AI & HPC, EMEA, Vast Data. “As HPC and AI converge, the most advanced supercomputing environments are being asked to support sensitive data, secure multi-tenancy, cloud-like services, new I/O patterns and governance at extreme scale. CSCS is helping prove that future first, showing why the data layer is becoming central to trusted, high-performance AI and HPC.”

The findings are significant for HPC centres evaluating how to support increasingly mixed workloads across simulation, AI training, data analytics and regulated research. By combining high-throughput performance with capabilities such as multi-tenancy, encryption, QoS, observability and non-disruptive operations, Vast gives supercomputing environments a data platform designed for both traditional HPC and emerging AI infrastructure requirements.

Vast will showcase its AI Operating System for AI and HPC infrastructure at ISC High Performance 2026, taking place June 23-25 in Hamburg, Germany. Attendees can meet Vast at booth G05 to learn more about its work with leading supercomputing centres and its role in powering the next generation of scientific computing, AI and trusted research environments. 

Both papers presented by CSCS at the Cray User Group will be available in the next months from the official website.