Programmable Fabrics with Optical Switches in AI Supercomputers
We explore the integration of Optical Circuit Switches (OCSs) in AI/HPC clusters' fabrics, to enhance resiliency in case of failures and enable dynamic topology reconfiguration for optimized deep-learning training, using a Layer-1 Software-Defined Network approach.
Authors
- Chen Avin (ORCID: https://orcid.org/0000-0002-6647-8002)
- Prethvi Kashinkunti
- Giannis Patronas
- Eitan Zahavi (ORCID: https://orcid.org/0000-0003-0679-7850)
- Dimitris Syrivelis (ORCID: https://orcid.org/0000-0002-2344-5665)
- Paraskevas Bakopoulos (ORCID: https://orcid.org/0000-0002-3635-1641)
- Elad Mentovich
- Julie Bernauer (ORCID: https://orcid.org/0000-0003-2345-995X)
- N. Argyris (ORCID: https://orcid.org/0000-0002-3079-6279)
- Zsolt-Alon Wertheimer
- Louis Capps
- Athanasios Fevgas
- Nikos Terzenidis
Institutions
- Nvidia (United States) (US)
Publication Details
- Published
- 2025-06-29
- DOI
- https://doi.org/10.23919/oecc/psc62146.2025.11109345
- Primary Topic
- Modular Robots and Swarm Intelligence
- Type
- article
- Field-Weighted Citation Impact
- 0.00