mlsanityData to 7 Oct 2026

Incast-Free MoE Rate-Based Scheduling

arXiv 2607.26340 ↗First posted 28 Jul 2026cs.NI cs.DC cs.LGcited by 0

On the map: Mixture-of-experts → Efficient MoE training and inference systems

Abstract (arXiv)

Mixture of Experts (MoE) architectures have become key to large language models; however, their typical round-robin (RR) scheduling introduces significant bottlenecks. In this paper, we demonstrate that RR causes a previously-undiscovered exponential incast phenomenon with MoE traffic. We propose an alternative proactive fair scheduling framework tailored for MoE workloads, which effectively prevents fabric oversubscription. We also outline how it can be implemented in NICs. Finally, through extensive simulations with real and synthetic workloads, we demonstrate that this framework consistently eliminates incast, maintains a near-100% link utilization, and reduces Collective Completion Time (CCT).

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Efficient MoE training and inference systems, one of 12 directions in Mixture-of-experts. Builds on 7 of its 29 references that are in the map. First posted on arXiv 28 Jul 2026.

Builds on

7 of 29 references.

Closest ancestors · 1

Same direction, newest first

  1. 2025FAST: An Efficient Scheduler for All-to-All GPU Communication14 May

Other references · 6

  1. 2022Demystifying the Nvidia Ampere Architecture through Microbenchmarking and Instruction-level Analysiscited by 9
  2. 2025InfiniteHBD: Building Datacenter-Scale High-Bandwidth Domain for LLM with Optical Circuit Switching Transceiverscited by 7
  3. 2024REPS: Recycled Entropy Packet Spraying for Adaptive Load Balancing and Failure Mitigationcited by 6
  4. 2025BigMac: A Communication-Efficient Mixture-of-Experts Model Structure for Fast Training and Inferencecited by 5
  5. 2024STrack: A Reliable Multipath Transport for AI/ML Clusterscited by 4
  6. 2026Dynamic Hierarchical Birkhoff-von Neumann Decomposition for All-to-All GPU Communicationcited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025MoE-Prism: Disentangling Monolithic Experts for Elastic MoE Services via Model-System Co-Designs
  2. 2025Advancing MoE Efficiency: A Collaboration-Constrained Routing (C2R) Strategy for Better Expert Parallelism Design
  3. 2025D2MoE: Dual Routing and Dynamic Scheduling for Efficient On-Device MoE-based LLM Serving
  4. 2026Coordinated Scheduling for MoE LLM Serving
  5. 2026Multi-Layer Scheduling for MoE-Based LLM Reasoning
  6. 2025From Tokens to Layers: Redefining Stall-Free Scheduling for MoE Serving with Layered Prefill
  7. 2025RailS: Load Balancing for All-to-All Communication in Distributed Mixture-of-Experts Training
  8. 2024MoE++: Accelerating Mixture-of-Experts Methods with Zero-Computation Experts