Hidden Convexity in Queueing Models

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Chen, Xin, Xin, Linwei, Zhao, Minda
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866915605951545344
author Chen, Xin
Xin, Linwei
Zhao, Minda
author_facet Chen, Xin
Xin, Linwei
Zhao, Minda
contents We study the joint control of arrival and service rates in queueing systems with the objective of minimizing long-run expected cost minus revenue. Although the objective function is non-convex, first-order methods have been empirically observed to converge to globally optimal solutions. This paper provides a theoretical foundation for this empirical phenomenon by characterizing the optimization landscape and identifying a hidden convexity: the problem admits a convex reformulation after an appropriate change of variables. Leveraging this hidden convexity, we establish the Polyak-Lojasiewicz-Kurdyka (PLK) condition for the original control problem, which excludes spurious local minima and ensures global convergence for first-order methods. Our analysis applies to a broad class of $GI/GI/1$ queueing models, including those with Gamma-distributed interarrival and service times. As a key ingredient in the proof, we establish a new convexity property of the expected queue length under a square-root transformation of the traffic intensity.
format Preprint
id arxiv_https___arxiv_org_abs_2511_03955
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Hidden Convexity in Queueing Models
Chen, Xin
Xin, Linwei
Zhao, Minda
Optimization and Control
Probability
We study the joint control of arrival and service rates in queueing systems with the objective of minimizing long-run expected cost minus revenue. Although the objective function is non-convex, first-order methods have been empirically observed to converge to globally optimal solutions. This paper provides a theoretical foundation for this empirical phenomenon by characterizing the optimization landscape and identifying a hidden convexity: the problem admits a convex reformulation after an appropriate change of variables. Leveraging this hidden convexity, we establish the Polyak-Lojasiewicz-Kurdyka (PLK) condition for the original control problem, which excludes spurious local minima and ensures global convergence for first-order methods. Our analysis applies to a broad class of $GI/GI/1$ queueing models, including those with Gamma-distributed interarrival and service times. As a key ingredient in the proof, we establish a new convexity property of the expected queue length under a square-root transformation of the traffic intensity.
title Hidden Convexity in Queueing Models
topic Optimization and Control
Probability
url https://arxiv.org/abs/2511.03955