Impact of RoCE Congestion Control Policies on Distributed Training of DNNs
Impact of RoCE Congestion Control Policies on Distributed Training of DNNs
Ahstract-RDMA over Converged Ethernet (RoCE) has gained significant attraction for datacenter networks due to its compatibility with conventional Ethernet-based fabric. However, the RDMA protocol is efficient only on (nearly) lossless networks, emphasizing the vital role of congestion control on RoCE networks. Unfortunately, the native RoCE congestion control scheme, based on …