Lightweight Latency Prediction Scheme for Edge Applications: A Rational Modelling Approach

2025

Read Lightweight Latency Prediction Scheme for Edge Applications: A Rational Modelling Approach (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-02501.txt.

Description

Mohan Liyanage, Eldiyar Zhantileuov, Ali Kadhum Idrees, and Rolf Schuster predict network delay for edge computing without pinging the network. Offloading tasks from a device to a nearby edge server only works in real time if the end-to-end latency is known in advance. The usual way to learn it, active probing, is itself intrusive. Their lightweight scheme instead builds a rational model, a ratio of polynomials, from passive features: frame size, arrival rate, and link utilisation.

Across extensive experiments with five-fold cross-validation, the model reached state-of-the-art prediction accuracy, reporting a mean absolute error of 0.0115 and an R-squared of 0.9847, while keeping inference fast. The authors frame the result as a strong trade-off between precision and efficiency against both traditional regression methods and neural networks, suitable for latency-sensitive edge applications that cannot afford heavyweight predictors.

Catalog record

YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE22 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-02501

Browse the full library · About the Tower of Babel