# FUSCO: High-Performance Distributed Data Shuffling via Transformation-Communication Fusion

arXiv: https://arxiv.org/abs/2512.22036 · HTML: https://mlsanity.com/paper/2512.22036/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Placed in Efficient MoE training and inference systems, one of 12 directions in Mixture-of-experts.
- First posted on arXiv 26 Dec 2025.

## Placed in

- Mixture-of-experts / [Efficient MoE training and inference systems](https://mlsanity.com/moe/efficient-moe-training-and-inference-systems/)

## Built on by (most cited first, 4 of 4)

- [Revealing the Challenges of Attention-FFN Disaggregation for Modern MoE Models and Hardware Systems](https://mlsanity.com/paper/2602.09721/) (2026)
- [FEPLB: Exploiting Copy Engines for Nearly Free MoE Load Balancing in Distributed Training](https://mlsanity.com/paper/2604.19654/) (2026)
- [Relay Buffer Independent Communication over Pooled HBM for Efficient MoE Inference on Ascend](https://mlsanity.com/paper/2605.06055/) (2026)
- [UBEP: Re-architecting Expert Parallelism Communication Library for Production Superpods](https://mlsanity.com/paper/2607.06202/) (2026)

## Abstract (arXiv)

Large-scale Mixture-of-Experts (MoE) models rely on expert parallelism for efficient training and inference, which splits experts across devices and necessitates distributed data shuffling to route each token to its assigned experts. However, existing communication libraries handle this shuffling poorly; its overhead can account for over half of end-to-end runtime. We present FUSCO, an MoE-friendly communication library that achieves efficient and lightweight data shuffling through fused data transformation and communication, based on the key observation that MoE's expert-major data layout conflicts with the device-major layout expected by communication operations. FUSCO captures the fine-grained data layout, which is then interpreted by a pipelined communication engine that performs the required shuffling efficiently along the communication path. Lightweight planning and load-balancing mechanisms complement the engine by eliminating redundant communication and dispersing traffic. Evaluations on representative benchmarks illustrate that FUSCO achieves up to 3.84× and 2.01× speedups over NCCL and DeepEP (the state-of-the-art MoE communication library), respectively. In end-to-end MoE tasks, compared to NCCL and DeepEP, FUSCO reduces the training latency by 1.17-1.39× and 1.10-1.19×, and lowers the first-token generation latency in inference by 1.09-1.25× and 1.06-1.16×.