Repository for Publications and Research Data
News from the ETH Library
Recently Added
Transitioning User and Identity Management for Alps
Item type: Conference Paper
Ceriani A.; Pagnamenta F.; Mazzoleni D.; et al. (2026)
The transition to Alps [4], a next-generation high-performance computing infrastructure at the Swiss National Supercomputing Centre (CSCS), required rethinking identity and access management (IAM) as a foundational architectural concern. Increasing platform heterogeneity, automation, and diversification of access lifecycles have made IAM a critical enabler for secure operation, delegation, and integration across modern HPC environments. This paper presents a modular IAM architecture that explicitly separates workflow initiation, authoritative identity and account lifecycle management, and backend service integration. Project- and resource-scoped workflows are initiated through a management platform, while identity authority, authentication, and lifecycle semantics are owned by dedicated backend services integrated with institutional IAM and accounting systems. We describe the architecture, its guiding principles, and its validation during the production transition to Alps. Particular attention is given to supporting heterogeneous identity types, enabling coexistence between legacy and new systems, and performing incremental migration without disruptive cut-overs. The paper also discusses operational lessons learned from running both systems concurrently during the transition. The architectural patterns and integration strategies described here may be applicable to other HPC centers facing similar evolution toward more automated and federated research infrastructures.
Architectural Isolation for Sensitive Workloads: Enabling Trusted Research Environments on HPE Cray EX Systems
Item type: Conference Paper
Witlox P.; Gila M.; Bonesana I.; et al. (2026)
The convergence of high-performance computing (HPC) with stringent data protection mandates - such as the U.S. Health Insurance Portability and Accountability Act (HIPAA), the Swiss revised Federal Act on Data Protection, and the European GDPR and European Health Data Space (EHDS) - presents a fundamental architectural challenge for shared supercomputing infrastructures. Traditionally, performance penalties associated with instruction-level encryption and traditional siloed small-scale infrastructure have rendered leadership-class systems impractical for sensitive biomedical research. This paper presents an architectural approach for extending Trusted Research Environments (TREs) as elastic data processing extensions onto Alps, an HPE Cray EX supercomputer at CSCS, enabling secure multi-tenant operation with near-native HPC performance. Our primary contribution is a control plane inversion model that positions the TRE as the authoritative policy source while treating the HPE Cray EX as a transparent execution substrate, un-siloing the TRE. We detail the integration of three core technologies: 1) hardware-enforced Layer-2 network isolation leveraging HPE Slingshot VLAN-based segmentation (with a strategic roadmap toward Virtual Network IDs (VNIs) for increased isolation); 2) tenant-managed storage encryption via the VAST Data Platform, and 3) a multi-layer "circuit breaker"mechanism that provides automated security enforcement across the infrastructure and workflow layers. Initial deployment with an open-source Kubernetes-based TRE platform developed by the Centre Hospitalier Universitaire Vaudois (CHUV) demonstrates that this architecture achieves near-native performance. By decoupling security concerns from the computational resources, we provide a scalable template for supercomputing centers to support high-stakes "Science for Society"use cases - such as neuroscience and personalized medicine - on shared, bare-metal HPC resources.
Sarus Suite: Cloud-native Containers for HPC
Item type: Conference Paper
Madonna A.; Chesi M.; Lee G.; et al. (2026)
High-performance computing (HPC) systems must support fast-moving software stacks, especially in AI/ML, while preserving scheduler control, scalable startup, and production performance. Yet many HPC container solutions rely on specialized runtime stacks that weaken continuity with mainstream cloud-native workflows and require ongoing effort to sustain compatibility with the evolving upstream ecosystem. We argue that HPC should specialize the integration layer while keeping the container engine aligned with upstream container evolution. We present Sarus Suite, an upstream-aligned HPC container architecture built around an unchanged Podman engine. Sarus Suite adds the HPC-specific functionality needed for production use through complementary system layers for declarative runtime specification, scheduler-native execution, scalable shared-image access, and standards-based host capability injection. We evaluate Sarus Suite on a Cray EX GH200 system using communication-intensive HPC workloads, large scale AI training, metadata-heavy startup workloads, and container startup measurements. Across PyFR, SPH-EXA, Megatron-LM, and Pynamic, Sarus Suite matches the performance and scaling of the production Enroot+Pyxis baseline while delivering consistently faster per-node container startup. The architecture also enables direct use of upstream OCI images, including NGC-based images, and supports cloud-native multi-container workflows expressed through Kubernetes manifests. These results show that HPC-grade containers do not require an HPC-specific runtime, provided that scheduler semantics, scalable image access, and host integration are implemented in explicit system layers. This preserves upstream continuity and software agility while maintaining scheduler control, scalability, and production performance.
Oxygen-dependent constraint of glycolysis through the mitochondrial electron transport chain
Item type: Journal Article
Guajardo-Grence A.; Fernández-Lamadrid M.; Moschini G.; et al. (2026)
Hypoxia-inducible factors (HIFs) are constitutively active in clear cell renal cell carcinoma (ccRCC) cells. Here, we investigated whether these cells could still respond to hypoxia independently of canonical HIF activation. We found that hypoxia provokes a NADH reductive stress that attenuates lactate release when the supply of extracellular pyruvate becomes limited, a response that was restored by the ectopic expression of the cytosolic and mitochondria-targeted NADH oxidase (LbNOX). Suppression of the mitochondrial electron transport chain (mtETC) alone is sufficient to attenuate glycolytic flux when extracellular pyruvate is limited, a situation consistent with metabolomic data from human ccRCCs. Furthermore, this response can be mimicked by pharmacological or genetic inhibition of mitochondrial complex IV (CIV) activity and restored by the mitochondrial-targeted alternative oxidase (AOX). Therefore, our data suggest that cells can respond to low oxygen tension through a mitochondrial-dependent elevation of NADH/NAD+, resulting in the suppression of glycolysis when pyruvate availability is compromised.
Coarse climate model grids underestimate coastal humid-heat extremes
Item type: Journal Article
Carniel C.E.; Wille J.D.; Brunner L.; et al. (2026)
Humid-heat extremes reach intensities close to the limit of human survivability around the Arabian Peninsula. Here, we analyze fully coupled kilometer-scale Icosahedral Nonhydrostatic (ICON) and Integrated Forecasting System-FinitevolumE Sea ice-Ocean Model (IFS-FESOM) simulations from the nextGEMS project, with original resolutions of around 9 kilometers, and compare them with station observations, ERA5, and CMIP6 projections. IFS-FESOM reproduces well the observed upper-tail wet-bulb temperature (TW) and lies near the high end of the CMIP6 distribution over coastal regions, while ICON simulates weaker extremes. Spatial smoothing suppresses localized coastal maxima, reducing daily-maximum TW extremes by up to 8°C and lowering exceedance frequencies across multiple thresholds. Native resolution fields also show stronger mid-century increases in exposure under SSP3-7.0 than their coarse-gridded counterparts. These results suggest that conventional coarse resolution projections may underestimate localized coastal humid-heat risk, although the magnitude of this effect remains model dependent.