HetCCL: Enabling Collective Communication For Mixed-Vendor Heterogeneous Clusters

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Wang, Yuejie, Chang, Tao, Zhao, Yuanyuan, Ao, Yulong, Gu, Zeyu, Li, Zhiyu, Jia, Yanmin, Zhang, Yan, Zhang, Mingjun, Liu, He, He, Yongzhe, Lin, Yonghua, Liu, Guyue
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911731542917120
author Wang, Yuejie
Chang, Tao
Zhao, Yuanyuan
Ao, Yulong
Gu, Zeyu
Li, Zhiyu
Jia, Yanmin
Zhang, Yan
Zhang, Mingjun
Liu, He
He, Yongzhe
Lin, Yonghua
Liu, Guyue
author_facet Wang, Yuejie
Chang, Tao
Zhao, Yuanyuan
Ao, Yulong
Gu, Zeyu
Li, Zhiyu
Jia, Yanmin
Zhang, Yan
Zhang, Mingjun
Liu, He
He, Yongzhe
Lin, Yonghua
Liu, Guyue
contents Training Large Language Models (LLMs) on heterogeneous clusters presents significant challenges for collective communication, as hardware from multiple vendors introduces diverse network and computational characteristics. Existing collective communication frameworks (e.g., NCCL, RCCL) designed for homogeneous environments fail to address mixed-hardware setups, while communication libraries with heterogeneous support (e.g., Gloo, OpenMPI) incur heavy overhead in the data path. This paper presents HetCCL, a framework that enables heterogeneous collective communication by efficient P2P transport across heterogeneous devices (e.g., GPUs), eliminating the host-device memory copy overhead while offloading the control to the CPUs. For combining collectives (e.g., AllReduce, ReduceScatter), HetCCL introduces a border-communicator mechanism that achieves vendor independence by using the intrinsic reduction in the combining collectives in vendor collective communication libraries. With efficient heterogeneous P2P transport and portable reduction mechanism, HetCCL proposes a hierarchical topology abstraction for heterogeneous clusters, dissecting collective communication into cluster-level primitives that guarantee optimal cross-cluster data transfer volume and optimal bandwidth utilization. We implement HetCCL with 4 different vendor support and evaluate it in 4 heterogeneous settings with benchmarks and end-to-end LLM tasks. Our evaluation shows that HetCCL achieves 17-19x higher bandwidth than Gloo in heterogeneous communications, and speeds up end-to-end training by up to 16.9% in the per-step-time.
format Preprint
id arxiv_https___arxiv_org_abs_2605_31000
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle HetCCL: Enabling Collective Communication For Mixed-Vendor Heterogeneous Clusters
Wang, Yuejie
Chang, Tao
Zhao, Yuanyuan
Ao, Yulong
Gu, Zeyu
Li, Zhiyu
Jia, Yanmin
Zhang, Yan
Zhang, Mingjun
Liu, He
He, Yongzhe
Lin, Yonghua
Liu, Guyue
Networking and Internet Architecture
Machine Learning
Training Large Language Models (LLMs) on heterogeneous clusters presents significant challenges for collective communication, as hardware from multiple vendors introduces diverse network and computational characteristics. Existing collective communication frameworks (e.g., NCCL, RCCL) designed for homogeneous environments fail to address mixed-hardware setups, while communication libraries with heterogeneous support (e.g., Gloo, OpenMPI) incur heavy overhead in the data path. This paper presents HetCCL, a framework that enables heterogeneous collective communication by efficient P2P transport across heterogeneous devices (e.g., GPUs), eliminating the host-device memory copy overhead while offloading the control to the CPUs. For combining collectives (e.g., AllReduce, ReduceScatter), HetCCL introduces a border-communicator mechanism that achieves vendor independence by using the intrinsic reduction in the combining collectives in vendor collective communication libraries. With efficient heterogeneous P2P transport and portable reduction mechanism, HetCCL proposes a hierarchical topology abstraction for heterogeneous clusters, dissecting collective communication into cluster-level primitives that guarantee optimal cross-cluster data transfer volume and optimal bandwidth utilization. We implement HetCCL with 4 different vendor support and evaluate it in 4 heterogeneous settings with benchmarks and end-to-end LLM tasks. Our evaluation shows that HetCCL achieves 17-19x higher bandwidth than Gloo in heterogeneous communications, and speeds up end-to-end training by up to 16.9% in the per-step-time.
title HetCCL: Enabling Collective Communication For Mixed-Vendor Heterogeneous Clusters
topic Networking and Internet Architecture
Machine Learning
url https://arxiv.org/abs/2605.31000