P4-based in-network aggregation system for Federated Learning that performs gradient summation directly inside programmable switches, reducing network congestion and server overhead while demonstrating the SwitchML paradigm using BMv2, Mininet, Scapy, and Docker.
p4 distributed-machine-learning federated-learning programmable-network in-network-computing in-network-aggregation switchml
-
Updated
Mar 12, 2026 - P4