Distributed deep learning (DDL) systems strongly depend on network performance. Current electronic packet switched (EPS) network architectures and technologies suffer from variable diameter topologies, low-bisection bandwidth and over-subscription affecting completion time of communication and collective operations. We introduce a near-exascale, full-bisection bandwidth, all-to-all, single-hop, all-optical network architecture with nanosecond reconfiguration called RAMP, which supports large-scale distributed and parallel computing systems (12.8~Tbps per node for up to 65,536 nodes). For the first time, a custom RAMP-x MPI strategy and a network transcoder is proposed to run MPI collective operations across the optical circuit switched (OCS) network in a schedule-less and contention-less manner. RAMP achieves 7.6-171× speed-up in completion time across all MPI operations compared to realistic EPS and OCS counterparts. It can also deliver a 1.3-16× and 7.8-58× reduction in Megatron and DLRM training time respectively} while offering 42-53× and 3.3-12.4× improvement in energy consumption and cost respectively.
@article{arxiv.2211.15226,
title = {RAMP: A Flat Nanosecond Optical Network and MPI Operations for Distributed Deep Learning Systems},
author = {Alessandro Ottino and Joshua Benjamin and Georgios Zervas},
journal= {arXiv preprint arXiv:2211.15226},
year = {2023}
}