Broadcast Rate Limits in Wi-Fi: A Forgotten Bottleneck for Collaborative Edge LLM Inference

2026-08-03

AI Summary

This work addresses the critical bottleneck imposed by the IEEE 802.11 standard’s 54 Mbps Wi-Fi broadcast rate limit in one-to-many communication for edge-based Mixture-of-Experts (MoE) large model inference. To overcome this constraint, the study introduces UDP broadcast into edge MoE inference for the first time, proposing a reliable transmission mechanism that combines timeout-based retransmission with out-of-order result collection. It further advocates expanding Wi-Fi broadcast from the control plane to the high-throughput data plane. Experimental results demonstrate a 1.4× speedup over NCCL and TCP implementations on an 8-node wired cluster. NS-3 simulations further validate the approach, achieving effective broadcast throughput equivalent to 64×, 43×, and 32× the standard limit at distances of 1 m, 2 m, and 5 m, respectively.

Abstract

LLM deployment is migrating from data centers to edge devices, where Mixture-of-Experts (MoE) models offer a promising path: sparse expert activation allows the model to be spread across multiple low-cost edge nodes. Distributed MoE inference repeatedly dispatches embeddings from one main node to many workers - a one-to-many pattern poorly served by the sequential unicasts of mainstream stacks (NCCL, TCP), yet naturally matched by UDP broadcast. We propose a UDP broadcast method for collaborative edge MoE inference, augmented with timeout-driven retransmission exploiting near deterministic latency in distributed MoE for reliability and unordered result gathering for robustness to expert mispredictions, yielding a consistent 1.4x speedup over NCCL and TCP on a wired 8-node cluster. In wireless settings, however, we uncover a deeper, long-forgotten bottleneck: IEEE 802.11 caps broadcast rates at 54 Mbps regardless of physical-layer capacity - a legacy policy built for sparse control traffic, not edge AI. NS-3 simulations at distances 1m, 2m and 5m show that the optimal rates are much higher (64x, 43x, and 32x, respectively) than the 54 Mbps cap applied in standard. Thus, we argue that broadcast is no longer a control-plane relic: it is time for Wi-Fi standards to treat it as a high-throughput data-plane citizen.

Problem

• broadcast rate limit
• Wi-Fi
• edge LLM inference
• Mixture-of-Experts
• IEEE 802.11

Innovation

• UDP broadcast
• Mixture-of-Experts
• edge LLM inference
• Wi-Fi broadcast rate limit
• distributed inference