Orchestrating Multimodal DNN Workloads in Wireless Neural Processing

TL;DR

This paper introduces a unified framework for optimizing multimodal DNN workloads in wireless neural processing, effectively reducing inference latency by overlapping wireless transmission with DNN execution.

Contribution

It proposes a novel end-to-end orchestration framework, O-WiN, with algorithms RTFS and PACS for improved scheduling and latency reduction in wireless neural processing.

Findings

PACS outperforms RTFS in high heterogeneity scenarios

Communication-computation pipelining accelerates multimodal DNN execution

Unified model enables effective workload orchestration in WNP

Abstract

In edge inference, wireless resource allocation and accelerator-level deep neural network (DNN) scheduling have yet to be co-optimized in an end-to-end manner. The lack of coordination between wireless transmission and accelerator-level DNN execution prevents efficient overlap, leading to higher end-to-end inference latency. To address this issue, this paper investigates multimodal DNN workload orchestration in wireless neural processing (WNP), a paradigm that integrates wireless transmission and multi-core accelerator execution into a unified end-to-end pipeline. First, we develop a unified communication-computation model for multimodal DNN execution and formulate the corresponding optimization problem. Second, we propose O-WiN, a framework that orchestrates DNN workloads in WNP through two tightly coupled stages: simulation-based optimization and runtime execution. Third, we develop two algorithms, RTFS and PACS. RTFS schedules communication and computation sequentially, whereas PACS interleaves them to enable pipeline parallelism by overlapping wireless data transfer with accelerator-level DNN execution. Simulation results demonstrate that PACS significantly outperforms RTFS under high modality heterogeneity by better masking wireless latency through communication-computation overlap, thereby highlighting the effectiveness of communication-computation pipelining in accelerating multimodal DNN execution in WNP.