A4 Vertaisarvioitu artikkeli konferenssijulkaisussa

Adaptive Workload Distribution for Accuracy-aware DNN Inference on Collaborative Edge Platforms




TekijätTaufique, Zain; Miele, Antonio; Liljeberg, Pasi; Kanduri, Anil

ToimittajaN/A

Konferenssin vakiintunut nimiAsia and South Pacific Design Automation Conference

Julkaisuvuosi2024

JournalProceedings of the Asia and South Pacific Design Automation Conference

Kokoomateoksen nimi2024 29th Asia and South Pacific Design Automation Conference (ASP-DAC)

Sarjan nimiProceedings of the Asia and South Pacific Design Automation Conference

Numero sarjassa29

Aloitussivu109

Lopetussivu114

ISBN979-8-3503-9355-2

eISBN979-8-3503-9354-5

ISSN2153-6961

eISSN2153-697X

DOIhttps://doi.org/10.1109/ASP-DAC58780.2024.10473987

Verkko-osoitehttps://ieeexplore.ieee.org/document/10473987

Preprintin osoitehttps://arxiv.org/pdf/2310.10157


Tiivistelmä

DNN inference can be accelerated by distributing the workload among a cluster of collaborative edge nodes. Heterogeneity among edge devices and accuracy-performance trade-offs of DNN models present a complex exploration space while catering to the inference performance requirements. In this work, we propose adaptive workload distribution for DNN inference, jointly considering node-level heterogeneity of edge devices, and application-specific accuracy and performance requirements. Our proposed approach combinatorially optimizes heterogeneity-aware workload partitioning and dynamic accuracy configuration of DNN models to ensure performance and accuracy guarantees. We tested our approach on an edge cluster of Odroid XU4, Raspberry Pi4, and Jetson Nano boards and achieved an average gain of 41.52\% in performance and 5.2\% in output accuracy as compared to state-of-the-art workload distribution strategies. 



Last updated on 2025-10-02 at 13:48