Bridge reduces All-to-All completion time by typically 3x to 10x and improves AllReduce by up to 6.6x over Ring by reusing optical subrings across multiple steps in reconfigurable networks.
Mars: Near-optimal throughput with shallow buffers in reconfigurable datacenter networks.Proc
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.NI 3years
2026 3roles
background 1polarities
background 1representative citing papers
Demand-aware topologies achieve at least 5/8 asymptotic throughput in the worst case, separating from the demand-oblivious bound of n/(2n-1) ≈ 1/2; computing optimal weak throughput is NP-hard while direct throughput is polynomial-time solvable.
A greedy max-weight decomposition strategy for MoE all-to-all communication on photonic fabrics improves overlap efficiency and reduces compute overheads compared to BvN by bounding the number of matchings.
citing papers explorer
-
Bridge: Optimizing Collective Communication Schedules in Reconfigurable Networks with Reusable Subrings
Bridge reduces All-to-All completion time by typically 3x to 10x and improves AllReduce by up to 6.6x over Ring by reusing optical subrings across multiple steps in reconfigurable networks.
-
A Separation Between Optimal Demand-Oblivious and Demand-Aware Network Throughput
Demand-aware topologies achieve at least 5/8 asymptotic throughput in the worst case, separating from the demand-oblivious bound of n/(2n-1) ≈ 1/2; computing optimal weak throughput is NP-hard while direct throughput is polynomial-time solvable.
-
Birkhoff Decompositions and Photonic Interconnects Wait! Don't Forget the Compute!
A greedy max-weight decomposition strategy for MoE all-to-all communication on photonic fabrics improves overlap efficiency and reduces compute overheads compared to BvN by bounding the number of matchings.