Further information about the HPC-SIG Hartree Meeting on the event page.
Research data movement: operator perspectives
James C. Womack (hybrid)
Bristol Centre for Supercomputing, University of Bristol
We are undertaking a project funded by the National Federated Compute Services (NFCS) NetworkPlus to investigate requirements and technologies for API-mediated research data transfer. Our goal is to survey communities, review current solutions, and evaluate specific technologies to propose what an effective API for data movement should provide. The perspective of HPC operators and infrastructure engineers is crucial to this analysis, particularly regarding examples of intra-facility data movement (e.g., between different storage infrastructure or tiers) which operators may be directly involved in.
We are therefore keen to engage with HPC operators and users at the meeting and propose a 30-minute hybrid session. The session would start with a 10-minute presentation introducing the NFCS NetworkPlus and our project. The remaining 20 minutes would be a facilitated discussion and requirements gathering session. We will engage the audience in discussions about data movement using a mix of interactive surveys (e.g., Slido, Mentimeter) and open discussion involving both in-person and online participants.
Feedback gathered will help us identify candidate technologies for further technical evaluation. Understanding to what extent existing solutions meet requirements will help us build a comprehensive picture of the current landscape and provide recommendations to NFCS roadmaps. We will share any slides presented alongside an anonymised summary of the discussion and requirements gathering session with the HPC-SIG community after the event.
Federated AI container platform: a feasibility study
Frederic Brochu-Williams (online)
University of Cambridge
A quick introduction to the half-day workshop following the main event and its related UKRI project: Feasibility Study on a federated, AI containerised platform/registry.
Isambard 3 processes and case studies from the University of Exeter
Jeremy Pike
University of Exeter
Isambard 3 is a Tier-2 HPC Facility hosted by the University of Bristol and operated by the GW4 Alliance. As a member of GW4 the University of Exeter has dedicated resource on Isambard 3 and local Research Software Engineer (RSE) support. We would like to briefly present our University of Exeter specific resource allocation and RSE support processes – ideally looking for feedback from the community. We will also highlight one or two case studies of live Isambard 3 projects.
Slurm Configuration for the Mary Coombs Cluster at the Hartree Centre
Mark Fellows
Hartree Centre, STFC
The talk will cover various configuration options we considered in order to satisfy requirements and, hopefully, maximise utilisation.
Configuring The Software Stack For MCC
Ronald Collins
Hartree Centre (STFC)
The Hartree Centre has recently taken delivery of a new HPC the Mary Coombs Cluster (MCC). We have chosen to make several changes in how we organise our software stack for users. Whilst Lmod is still the primary focus we have also utilised spack environments for the more time consuming installs. But we have also attempted to keep as much control of our software stack as possible, for multiple reasons which will be discussed.
An Update on High Performance Computing at the University of Southampton
Aaron Parsons
University of Southampton
The University of Southampton is currently a hive of High Performance Computing activity, offering researchers the use of 3 high performance computing clusters, one of which is a home-built system. This talk will provide an update on current activities and challenges, and a look to the future of computing at Southampton and how things are set to change over the next year as we welcome Iridis 7: a high performance cluster targeted at accelerated and AI applications.
Attaching enterprise class VMs to CLOS network leaf switches from a centralised cluster
Jonathan Churchill
STFC RAL
Historically our enterprise virtualisation clusters were deployed to large layer2 networks where all the VMs were on the same networks and increasingly not on the networks where the HPC’s were. This is increasingly an issue with large 3/5 tier “CLOS” leaf/spine network topologies. During our enterprise virtualisation refresh at RAL we’ve taken the opportunity to build a single cluster that allows us to attach enterprise VMs to any L2 network segment anywhere in our infrastructure using a combination of EVPN/VXLAN and more traditional approaches.
Testing Virtualised MIG backed GPUs on our HPC
Sam Gallop (online)
Norwich Bioscience Institute
At the NBI we’re working towards a hybrid approach of physical and cloud HPC. This talk goes over our work to partition our GPUs and run them across isolated VMs in our HPC and cloud infrastructure.
Quantum-Centric Supercomputing at The Hartree Centre
Dr Stefano Mensa – Advanced Computing and Emerging Technologies Lead
Hartree Centre
TBC
The History of Computing at Daresbury Laboratory
Rob Allan
STFC Hartree Centre
We will outline the history of computing at Daresbury from the early IBM 1800, through the Cray-1 years with the formation of the Computational Science Department to the HPCx service and Hartree Centre systems.
Completely powering off a 38MW data centre – what could possibly go wrong…
Paul Clark
EPCC
For the first time in over 25 years, the Advanced Computing Facility near Edinburgh was powered down- completely. This talk gives an overview of the approach, challenges, logistics and risks of doing so.
Federated Access Management for UK DRIs: Building a Community of Practice
Gwen Dawes (hybrid)
University of Cambridge
Federated Identity and Access Management (IAM) is rapidly becoming a critical requirement for UK research computing, enabling secure, seamless access across institutions, platforms, and communities. With the NFCS Network Plus funding a new Community of Practice (CoP) around federated IAM, there is now an opportunity to share approaches and air challenges and victories in adoption and interoperability. This panel session aims to bring together voices from across the community, including representatives from national initiatives and institutional deployments to share practical experiences with solutions such as MyAccessID, Keycloak etc. Through a structured discussion and open Q&A, participants will also be invited to contribute their own perspectives on pain points, success stories, and priorities for the CoP. The session will highlight how federated IAM aligns with NFCS roadmaps, gather input on the needs of the HPC community. Confirmed panel members: Christopher Woods (Bristol) – MyAccessID and Isambard-AI Jisc representative – national perspective on eduGAIN