HybridEP: Scaling Expert Parallelism to Cross-Datacenter Scenario via Hybrid Expert/Data Transmission

TL;DR

HybridEP introduces a dynamic, model-guided framework for expert parallelism in MoE models, significantly improving scalability and training speed across multiple data centers with limited bandwidth.

Contribution

It proposes a novel hybrid expert/data transmission approach with a stream-based model and topology optimization techniques to enhance cross-DC MoE training scalability.

Findings

HybridEP outperforms state-of-the-art systems by up to 5.6x under bandwidth constraints.

Achieves up to 1.45x speedup with 1000 data centers in simulations.

Effectively reduces communication overhead in low-bandwidth, cross-DC MoE training.

Abstract

Mixture-of-Experts (MoE) has become a popular architecture for scaling large models. However, the rapidly growing scale outpaces model training on a single DC, driving a shift toward a more flexible, cross-DC training paradigm. Under this, Expert Parallelism (EP) of MoE faces significant scalability issues due to the limited cross-DC bandwidth. Specifically, existing EP optimizations attempt to overlap data communication and computation, which has little benefit in low-bandwidth scenarios due to a much longer data communication time. Therefore, the trends of cross-DC EP scaling is fast becoming a critical roadblock to the continued growth of MoE models. To address this, we propose HybridEP, a modeling-guided framework to optimize EP under constrained bandwidth. Our key idea is to dynamically transform the spatial placement of experts to reduce data communication traffic and frequency, thereby minimizing EP's communication overheads. However, it is non-trivial to find the optimal solution because it complicates the original communication pattern by mixing data and expert communication. We therefore build a stream-based model to determine the optimal transmission ratio. Guided by this, we incorporate two techniques: (1) domain-based partition to construct the mapping between hybrid patterns and specific communication topology at GPU level, and (2) parameter-efficient migration to further refine this topology by reducing expert transmission overhead and enlarging the domain size. Combining all these designs, HybridEP can be considered as a more general EP with better scalability. Experimental results show that HybridEP outperforms existing state-of-the-art MoE training systems by up to 5.6x under constrained bandwidth. We further compare HybridEP and EP on large-scale simulations. HybridEP achieves up to 1.45x speedup with 1k DCs under different bandwidths.