Preprint
Oct 2026
Ofan: Optimal Load Balancing for AI Training
The extreme collective completion time (CCT) demands of AI workloads challenge existing packet spraying algorithms, which can have trouble efficiently load-balancing workloads that are sent at full line rates. We trace this to a structural cause: on a fat tree, once a packet picks its upward path, the downward path to...
Sarah McClure, E. Cohen, Jakob Krebs et al.
· 0 citations