Interim Supercomputing Service

Interim Supercomputing Service
Name	Interim Supercomputing Service
Type	Computing service
Established	2020s
Headquarters	Undisclosed / Distributed
Services	High-performance computing, data analytics, cloud bursting

Interim Supercomputing Service

The Interim Supercomputing Service is a transitional high-performance computing provision designed to bridge capacity gaps between legacy facilities and next-generation exascale installations. It provides temporary compute, storage, and networking resources to researchers, laboratories, agencies, and private firms while facilitating migration to permanent platforms. The service interacts with regional centers, national laboratories, international consortia, and vendor ecosystems to deliver compute cycles and specialist software stacks.

Overview

The Interim Supercomputing Service operates as an ad hoc consortium involving stakeholders such as Argonne National Laboratory, Lawrence Berkeley National Laboratory, Oak Ridge National Laboratory, National Aeronautics and Space Administration, European Organization for Nuclear Research, National Science Foundation, and private vendors like NVIDIA, AMD, Intel, Google, and Amazon Web Services. It supplies temporary allocations for projects tied to institutions such as Massachusetts Institute of Technology, Stanford University, University of California, Berkeley, California Institute of Technology, Princeton University, and international partners including CERN, Max Planck Society, École Polytechnique Fédérale de Lausanne, and RIKEN. Typical users include teams from NASA Jet Propulsion Laboratory, IBM, Microsoft Research, Los Alamos National Laboratory, Sandia National Laboratories, and non-profits like The Alan Turing Institute.

History and Origins

Origins trace to coordination efforts after capacity shortfalls at facilities like Cray Research installations and procurement delays for exascale systems following procurement cycles involving vendors such as Hewlett Packard Enterprise and Fujitsu. Early models borrowed governance elements from consortia such as XSEDE and PRACE, and drew on lessons from transitions like the procurement of Summit and Fugaku. Policy drivers included funding programs overseen by U.S. Department of Energy, directives influenced by legislative acts debated in bodies like the United States Congress and coordination with agencies including European Commission initiatives. Stakeholder meetings involved representatives from National Institutes of Health, Defense Advanced Research Projects Agency, Wellcome Trust, and philanthropic organizations such as Gordon and Betty Moore Foundation.

Infrastructure and Architecture

The service leverages a heterogeneous architecture combining accelerators from NVIDIA and AMD, CPUs from Intel and ARM partners, and interconnects using technologies from Mellanox Technologies and standards like InfiniBand and Ethernet. Storage layers integrate systems from NetApp, Dell Technologies, and parallel file systems such as Lustre and BeeGFS, while orchestration uses software by Kubernetes foundations, job schedulers like Slurm Workload Manager, and middleware rooted in projects from OpenStack and Apache Mesos. Security, authentication, and identity federation incorporate services from InCommon Federation, OpenID Foundation, and practices influenced by National Institute of Standards and Technology frameworks. Data management often interoperates with archives at National Center for Supercomputing Applications, Pawsey Supercomputing Centre, and Swiss National Supercomputing Centre.

Services and Access Models

Access models range from allocation-based peer review systems similar to PRACE and XSEDE awards, to commercial procurement via providers like Amazon Web Services and Google Cloud Platform, and rapid-response emergency allocations coordinated with agencies such as National Oceanic and Atmospheric Administration and European Space Agency. Services include batch HPC allocations, interactive notebooks developed with tools from Jupyter Project, container registries influenced by Docker, and reproducible workflows enabled by CWL and Nextflow. Training and community engagement often partner with academic programs at University of Cambridge, Imperial College London, University of Toronto, and initiatives led by IEEE and ACM.

Use Cases and Applications

The Interim Supercomputing Service supports computational fluid dynamics work tied to teams at Boeing and Airbus, climate modeling for groups at Met Office and European Centre for Medium-Range Weather Forecasts, genomics and bioinformatics projects associated with Wellcome Sanger Institute and Broad Institute, and machine learning workloads from research labs at DeepMind and OpenAI. Other applications include particle physics simulations for collaborations at CERN, materials science governed by research at Argonne National Laboratory (ANL), and epidemiological modeling supported by public health groups such as Centers for Disease Control and Prevention and World Health Organization partners.

Governance, Funding, and Partnerships

Governance models blend principles from consortia such as XSEDE, PRACE, and international partnerships like Partnership for Advanced Computing in Europe arrangements. Funding sources combine appropriations from agencies like U.S. Department of Energy, grants from National Science Foundation, contracts with corporations including Microsoft and IBM, and contributions from philanthropic entities such as Wellcome Trust. Partnerships span national laboratories, academic institutions, cloud providers, and standards organizations including Open Grid Forum and The Linux Foundation.

Challenges and Future Developments

Challenges include supply chain constraints experienced by vendors like NVIDIA and Intel, software portability across architectures championed by projects at Khronos Group and OpenMP, and long-term sustainability issues seen in transitions from systems such as Titan to successor platforms. Future developments emphasize integration with exascale systems like Frontier and Aurora, stronger links to international initiatives such as EuroHPC Joint Undertaking, and adoption of technologies from emerging vendors including Graphcore and Cerebras Systems. Ongoing policy discussions involve procurement frameworks in bodies like European Parliament and budget allocations debated in United States Congress committees.

Category:Supercomputing services