InferEdge: characterizing edge inference performance

Gabriel Orion Kai Antar; Blesson Varghese

doi:10.1145/3773274.3774257

InferEdge: characterizing edge inference performance

Gabriel Orion Kai Antar, Blesson Varghese^*

^*Corresponding author for this work

School of Computer Science

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

Machine learning (ML) inference is a key workload on edge computing resources, facilitated by deployment mechanisms such as Docker containers or WebAssembly. However, the performance characteristics of the deployment mechanisms for edge inference workloads across different target hardware platforms remain unknown. Therefore, this paper introduces InferEdge, a software suite that automates edge inference performance characterization for a range of edge inference models on heterogeneous processors. Experimental results obtained demonstrate that the suitability of each deployment mechanism depends on the workload and target platform, validating the need for a suite that can automate performance characterization and offer insights. InferEdge is available from https://github.com/blessonvar/InferEdge.

Original language	English
Title of host publication	Proceedings of the IEEE/ACM international conference on utility and cloud computing (UCC 2025)
Place of Publication	New York
Publisher	ACM
Number of pages	10
ISBN (Electronic)	9798400722851
DOIs	https://doi.org/10.1145/3773274.3774257
Publication status	Published - 31 Dec 2025
Event	18th IEEE/ACM International Conference on Utility and Cloud Computing - Nantes University, Nantes, France Duration: 1 Dec 2025 → 4 Dec 2025 https://ucc2025.gitlabpages.inria.fr/web/

Conference

Conference	18th IEEE/ACM International Conference on Utility and Cloud Computing
Abbreviated title	UCC2025
Country/Territory	France
City	Nantes
Period	1/12/25 → 4/12/25
Internet address	https://ucc2025.gitlabpages.inria.fr/web/

Keywords

Edge inference
Performance characterization
WebAssembly
Docker
Edge computing

Access to Document

10.1145/3773274.3774257

Antar_2025_UCC_InferEdge-edge-inference_AAM
Copyright © 2025 ACM / the Author(s). This work has been made available online in accordance with publisher policies or with permission. Permission for further reuse of this content should be sought from the publisher or the rights holder. This is the author created accepted manuscript following peer review and may differ slightly from the final published version. The final published version of this work is available at https://doi.org/10.1145/3773274.3774257
Accepted author manuscript, 1.15 MB

Cite this

@inproceedings{5a608433932b41c68da4958bb9573ec9,

title = "InferEdge: characterizing edge inference performance",

abstract = "Machine learning (ML) inference is a key workload on edge computing resources, facilitated by deployment mechanisms such as Docker containers or WebAssembly. However, the performance characteristics of the deployment mechanisms for edge inference workloads across different target hardware platforms remain unknown. Therefore, this paper introduces InferEdge, a software suite that automates edge inference performance characterization for a range of edge inference models on heterogeneous processors. Experimental results obtained demonstrate that the suitability of each deployment mechanism depends on the workload and target platform, validating the need for a suite that can automate performance characterization and offer insights. InferEdge is available from https://github.com/blessonvar/InferEdge.",

keywords = "Edge inference, Performance characterization, WebAssembly, Docker, Edge computing",

author = "Antar, \{Gabriel Orion Kai\} and Blesson Varghese",

note = "Funding: This research is supported by the UK Research and Innovation grant EP/Y028813/1.; 18th IEEE/ACM International Conference on Utility and Cloud Computing, UCC2025 ; Conference date: 01-12-2025 Through 04-12-2025",

year = "2025",

month = dec,

day = "31",

doi = "10.1145/3773274.3774257",

language = "English",

booktitle = "Proceedings of the IEEE/ACM international conference on utility and cloud computing (UCC 2025)",

publisher = "ACM",

address = "United States",

url = "https://ucc2025.gitlabpages.inria.fr/web/",

}

TY - GEN

T1 - InferEdge

T2 - 18th IEEE/ACM International Conference on Utility and Cloud Computing

AU - Antar, Gabriel Orion Kai

AU - Varghese, Blesson

N1 - Funding: This research is supported by the UK Research and Innovation grant EP/Y028813/1.

PY - 2025/12/31

Y1 - 2025/12/31

N2 - Machine learning (ML) inference is a key workload on edge computing resources, facilitated by deployment mechanisms such as Docker containers or WebAssembly. However, the performance characteristics of the deployment mechanisms for edge inference workloads across different target hardware platforms remain unknown. Therefore, this paper introduces InferEdge, a software suite that automates edge inference performance characterization for a range of edge inference models on heterogeneous processors. Experimental results obtained demonstrate that the suitability of each deployment mechanism depends on the workload and target platform, validating the need for a suite that can automate performance characterization and offer insights. InferEdge is available from https://github.com/blessonvar/InferEdge.

AB - Machine learning (ML) inference is a key workload on edge computing resources, facilitated by deployment mechanisms such as Docker containers or WebAssembly. However, the performance characteristics of the deployment mechanisms for edge inference workloads across different target hardware platforms remain unknown. Therefore, this paper introduces InferEdge, a software suite that automates edge inference performance characterization for a range of edge inference models on heterogeneous processors. Experimental results obtained demonstrate that the suitability of each deployment mechanism depends on the workload and target platform, validating the need for a suite that can automate performance characterization and offer insights. InferEdge is available from https://github.com/blessonvar/InferEdge.

KW - Edge inference

KW - Performance characterization

KW - WebAssembly

KW - Docker

KW - Edge computing

UR - https://doi.org/10.1145/3773274

U2 - 10.1145/3773274.3774257

DO - 10.1145/3773274.3774257

M3 - Conference contribution

BT - Proceedings of the IEEE/ACM international conference on utility and cloud computing (UCC 2025)

PB - ACM

CY - New York

Y2 - 1 December 2025 through 4 December 2025

ER -

InferEdge: characterizing edge inference performance

Abstract

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this