Abstract
Existing approaches for managing congestion in an artificial intelligence (AI) data center (DC) network fabric typically involve dedicated hardware and are often reactive in nature. Proposed herein are techniques that can be utilized to avoid fabric congestion instead of building reactive techniques with new hardware. Specifically, a Segment Routing-Traffic Engineering (SR-TE) agent is proposed herein that can be utilized to avoid fabric congestion, which can enable enterprises to use their existing data centers for AI training.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Ali, Zafar; Filsfils, Clarence; Clad, Francois; Camarillo, Pablo; and McDougall, Bruce, "AVOIDING FABRIC CONGESTION USING AN SR-TE AGENT FOR AI TRAINING USING EXISTING DATA CENTERS", Technical Disclosure Commons, (February 03, 2026)
https://www.tdcommons.org/dpubs_series/9266