Chronos: Prescheduled circuit switching for LLM training
Hundreds of thousands of accelerators are used to train LLMs, with accelerators connected by packet-switched AI fabrics. In this paper, we ask if the fabric can be built entirely from time-synchronized circuit switches instead. The goal would be to reduce power, increase switching capacity, or reduce the number of network tiers.
Authors
- Nick McKeown (ORCID: https://orcid.org/0009-0000-3657-3933)
- Sundararajan Renganathan (ORCID: https://orcid.org/0009-0008-7853-9440)
Institutions
- Stanford University (US)
Publication Details
- Published
- 2025-09-02
- DOI
- https://doi.org/10.1145/3748273.3749210
- Primary Topic
- Parallel Computing and Optimization Techniques
- Type
- article
- Field-Weighted Citation Impact
- 0.00