In-Network Aggregation with Transport Transparency for Distributed Training

Shuo Liu, Qiaoling Wang, Junyi Zhang, Wenfei Wu, Qinliang Lin, Yao Liu, Meng Xu, Marco Canini, Ray C.C. Cheung, Jianfei He

Research output: Chapter in Book/Report/Conference proceedingConference contribution

6 Scopus citations

Abstract

Recent In-Network Aggregation (INA) solutions offload the all-reduce operation onto network switches to accelerate and scale distributed training (DT). On end hosts, these solutions build custom network stacks to replace the transport layer. The INA-oriented network stack cannot take advantage of the state-of-the-art performant transport layer implementation, and also causes complexity in system development and operation. We design a transport-transparent INA primitive named NetReduce for modern multi-rack data centers. NetReduce runs beneath the transport layer. The switch performs aggregation operations but preservesdata transmission connections. The host uses RoCE as its transport layer to deliver gradient messages and receive aggregation results. NetReduce achieves performance gains from both INA and RoCE: linear scalability, traffic reduction, and bandwidth freeing-up from INA-high throughput, low latency, and low CPU overhead from RoCE. For jobs spanning several multi-GPU machines, we also devise parallel all-reduce based on NetReduce to make use of intra-machine and inter-machine bandwidth efficiently. We prototype NetReduce on an FPGA board attached to an Ethernet switch. We compare NetReduce with existing programmable switch-based solutions and justify the FPGA-based design choice. We evaluate NetReduce's performance by training typical Deep Neural Network models on single-GPU and multi-GPU testbeds. NetReduce inter-operates with the existing Ethernet transport layer, is training-framework friendly, accelerates network-intensive DT jobs effectively (e.g., 70% for AlexNet), reduces CPU overheads (e.g., only one core for transmission), and is cost-effective (e.g., only 2.40% more capital expense and 0.68% more power consumption making 12.3-57.9% more performance acceleration).
Original languageEnglish (US)
Title of host publicationProceedings of the 28th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 3
PublisherACM
Pages376-391
Number of pages16
ISBN (Print)9781450399180
DOIs
StatePublished - Mar 25 2023

Bibliographical note

KAUST Repository Item: Exported on 2023-05-26
Acknowledgements: We thank our shepherd and all the anonymous reviewers for their helpful feedback. Part of the evaluation in this work is done by using the resources of Peng Cheng Laboratory.

Fingerprint

Dive into the research topics of 'In-Network Aggregation with Transport Transparency for Distributed Training'. Together they form a unique fingerprint.

Cite this