Geometric lower bounds for the steady-state occupancy of processing networks with limited connectivity

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Goldsztajn, Diego, Ferragut, Andres
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909001018507264
author Goldsztajn, Diego
Ferragut, Andres
author_facet Goldsztajn, Diego
Ferragut, Andres
contents We consider processing networks where multiple dispatchers are connected to single-server queues by a bipartite compatibility graph, modeling constraints that are common in data centers and cloud networks due to geographic reasons or data locality issues. We prove lower bounds for the steady-state occupancy, i.e., the complementary cumulative distribution function of the empirical queue length distribution. The lower bounds are geometric with ratios given by two flexibility metrics: the average degree of the dispatchers and a novel metric that averages the minimum degree over the compatible dispatchers across the servers. Using these lower bounds, we establish that the asymptotic performance of a growing processing network cannot match that of the classic Power-of-$d$ or JSQ policies unless the flexibility metrics approach infinity in the large-scale limit.
format Preprint
id arxiv_https___arxiv_org_abs_2505_08974
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Geometric lower bounds for the steady-state occupancy of processing networks with limited connectivity
Goldsztajn, Diego
Ferragut, Andres
Probability
Performance
60K25 (Primary) 68M20 (Secondary)
We consider processing networks where multiple dispatchers are connected to single-server queues by a bipartite compatibility graph, modeling constraints that are common in data centers and cloud networks due to geographic reasons or data locality issues. We prove lower bounds for the steady-state occupancy, i.e., the complementary cumulative distribution function of the empirical queue length distribution. The lower bounds are geometric with ratios given by two flexibility metrics: the average degree of the dispatchers and a novel metric that averages the minimum degree over the compatible dispatchers across the servers. Using these lower bounds, we establish that the asymptotic performance of a growing processing network cannot match that of the classic Power-of-$d$ or JSQ policies unless the flexibility metrics approach infinity in the large-scale limit.
title Geometric lower bounds for the steady-state occupancy of processing networks with limited connectivity
topic Probability
Performance
60K25 (Primary) 68M20 (Secondary)
url https://arxiv.org/abs/2505.08974