FedRDMA: Communication-Efficient Cross-Silo Federated LLM via Chunked RDMA Transmission

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Zeling, Cai, Dongqi, Zhang, Yiran, Xu, Mengwei, Wang, Shangguang, Zhou, Ao
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917602518892544
author Zhang, Zeling
Cai, Dongqi
Zhang, Yiran
Xu, Mengwei
Wang, Shangguang
Zhou, Ao
author_facet Zhang, Zeling
Cai, Dongqi
Zhang, Yiran
Xu, Mengwei
Wang, Shangguang
Zhou, Ao
contents Communication overhead is a significant bottleneck in federated learning (FL), which has been exaggerated with the increasing size of AI models. In this paper, we propose FedRDMA, a communication-efficient cross-silo FL system that integrates RDMA into the FL communication protocol. To overcome the limitations of RDMA in wide-area networks (WANs), FedRDMA divides the updated model into chunks and designs a series of optimization techniques to improve the efficiency and robustness of RDMA-based communication. We implement FedRDMA atop the industrial federated learning framework and evaluate it on a real-world cross-silo FL scenario. The experimental results show that \sys can achieve up to 3.8$\times$ speedup in communication efficiency compared to traditional TCP/IP-based FL systems.
format Preprint
id arxiv_https___arxiv_org_abs_2403_00881
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle FedRDMA: Communication-Efficient Cross-Silo Federated LLM via Chunked RDMA Transmission
Zhang, Zeling
Cai, Dongqi
Zhang, Yiran
Xu, Mengwei
Wang, Shangguang
Zhou, Ao
Machine Learning
Distributed, Parallel, and Cluster Computing
Networking and Internet Architecture
Communication overhead is a significant bottleneck in federated learning (FL), which has been exaggerated with the increasing size of AI models. In this paper, we propose FedRDMA, a communication-efficient cross-silo FL system that integrates RDMA into the FL communication protocol. To overcome the limitations of RDMA in wide-area networks (WANs), FedRDMA divides the updated model into chunks and designs a series of optimization techniques to improve the efficiency and robustness of RDMA-based communication. We implement FedRDMA atop the industrial federated learning framework and evaluate it on a real-world cross-silo FL scenario. The experimental results show that \sys can achieve up to 3.8$\times$ speedup in communication efficiency compared to traditional TCP/IP-based FL systems.
title FedRDMA: Communication-Efficient Cross-Silo Federated LLM via Chunked RDMA Transmission
topic Machine Learning
Distributed, Parallel, and Cluster Computing
Networking and Internet Architecture
url https://arxiv.org/abs/2403.00881