MoE-Hub: Taming Software Complexity for Seamless MoE Overlap with Hardware-Accelerated Communication on Multi-GPU Systems

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Zhou, Zhuoshan, Zhang, Chen, Zhang, Shuyi, Zhang, Qijun, Wang, Haibo, Zhou, Zhe, Tu, Zhipeng, Sun, Guangyu, Diao, Yijia, Ji, Zhigang, Leng, Jingwen, He, Guanghui, Guo, Minyi
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915987515768832
author Zhou, Zhuoshan
Zhang, Chen
Zhang, Shuyi
Zhang, Qijun
Wang, Haibo
Zhou, Zhe
Tu, Zhipeng
Sun, Guangyu
Diao, Yijia
Ji, Zhigang
Leng, Jingwen
He, Guanghui
Guo, Minyi
author_facet Zhou, Zhuoshan
Zhang, Chen
Zhang, Shuyi
Zhang, Qijun
Wang, Haibo
Zhou, Zhe
Tu, Zhipeng
Sun, Guangyu
Diao, Yijia
Ji, Zhigang
Leng, Jingwen
He, Guanghui
Guo, Minyi
contents The Mixture-of-Experts (MoE) architecture is crucial for scaling large language models, but its scalability is severely limited by inter-GPU communication bottlenecks in multi-GPU systems. Although overlapping communication with computation is a widely recognized optimization, its effective deployment still remains challenging, both in terms of performance and programmability. In this work, we identify the root cause as a fundamental abstraction mismatch between MoE's dynamic, irregular token-to-expert mapping and the static, address-centric communication model of modern GPUs, which necessitates a complex software mediation phase to resolve addresses before data transfers, limiting performance and software flexibility. To resolve this, we propose MoE-Hub, a hardware-software co-design that introduces a destination-agnostic communication paradigm. MoE-Hub decouples data transmission from address management, allowing producers to send data immediately after routing using only a logical destination, while address allocation and data-flow orchestration are handled transparently by lightweight hardware in the GPU hub. By hardware-accelerating the entire communication control plane, MoE-Hub enables seamless and transparent overlap. Our evaluation shows that MoE-Hub achieves 1.40x-3.08x per-layer and 1.21x-1.98x end-to-end speedup over state-of-the-art systems.
format Preprint
id arxiv_https___arxiv_org_abs_2605_05888
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle MoE-Hub: Taming Software Complexity for Seamless MoE Overlap with Hardware-Accelerated Communication on Multi-GPU Systems
Zhou, Zhuoshan
Zhang, Chen
Zhang, Shuyi
Zhang, Qijun
Wang, Haibo
Zhou, Zhe
Tu, Zhipeng
Sun, Guangyu
Diao, Yijia
Ji, Zhigang
Leng, Jingwen
He, Guanghui
Guo, Minyi
Hardware Architecture
Distributed, Parallel, and Cluster Computing
The Mixture-of-Experts (MoE) architecture is crucial for scaling large language models, but its scalability is severely limited by inter-GPU communication bottlenecks in multi-GPU systems. Although overlapping communication with computation is a widely recognized optimization, its effective deployment still remains challenging, both in terms of performance and programmability. In this work, we identify the root cause as a fundamental abstraction mismatch between MoE's dynamic, irregular token-to-expert mapping and the static, address-centric communication model of modern GPUs, which necessitates a complex software mediation phase to resolve addresses before data transfers, limiting performance and software flexibility. To resolve this, we propose MoE-Hub, a hardware-software co-design that introduces a destination-agnostic communication paradigm. MoE-Hub decouples data transmission from address management, allowing producers to send data immediately after routing using only a logical destination, while address allocation and data-flow orchestration are handled transparently by lightweight hardware in the GPU hub. By hardware-accelerating the entire communication control plane, MoE-Hub enables seamless and transparent overlap. Our evaluation shows that MoE-Hub achieves 1.40x-3.08x per-layer and 1.21x-1.98x end-to-end speedup over state-of-the-art systems.
title MoE-Hub: Taming Software Complexity for Seamless MoE Overlap with Hardware-Accelerated Communication on Multi-GPU Systems
topic Hardware Architecture
Distributed, Parallel, and Cluster Computing
url https://arxiv.org/abs/2605.05888