Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Resource Allocation in Mobile Networks: A Decision Model Of Jockeying in Queues

Anthony Kiggundu, Bin Han, Dennis Krummacker, Hans D. Schotten

TL;DR

The paper tackles decentralized resource allocation for multi-tenant network slicing in next-generation networks by modeling impatient tenants' jockeying between queues as a behavioral decision process. It develops a two-queue $M/M/C$ system with join-the-shorter-queue admissions, a Bayes-based mechanism for arrival probabilities, and Monte Carlo simulations to study jockeying frequency. A key theoretical result is the derived expected jockeying frequency $E[\xi] = d\,\mu_j/(\mu_i+\mu_j)$, validated through simulations and used to illuminate how waiting-time differences and service-rate heterogeneity drive switching behavior. The work contributes a decentralized perspective on queue management in network slices, accompanied by an open-source toolkit for simulating behavioral queuing in MEC/6G contexts, with implications for designing slice-aware resource sharing. Overall, it provides actionable insights into when jockeying reduces individual delays and how to balance local gains against system-wide performance in decentralized mobile networks.

Abstract

Use-case-specific network slicing in decentralized multi-tenancy cloud environments is a promising approach to bridge the gap between the demand and supply of resources in next-generation communication networks. Our findings associate different slice profiles to queues in a multi-server setting, such that tenants continuously assess their preferences and make rational decisions to minimize the queuing delay. Deviated from classical approaches that statistically model the jockeying phenomena in queuing systems, our work pioneers to setup a behavioral model of jockeying impatient tenants. This will serve as a basis for decentralized management of multi-queue systems, where the decision to jockey is individually made by each tenant upon its up-to-date assessment of expected waiting time. Additionally, we carry out numerical simulations to empirically unravel the parametric dependencies of the tenants' jockeying behavior.

Resource Allocation in Mobile Networks: A Decision Model Of Jockeying in Queues

TL;DR

The paper tackles decentralized resource allocation for multi-tenant network slicing in next-generation networks by modeling impatient tenants' jockeying between queues as a behavioral decision process. It develops a two-queue system with join-the-shorter-queue admissions, a Bayes-based mechanism for arrival probabilities, and Monte Carlo simulations to study jockeying frequency. A key theoretical result is the derived expected jockeying frequency , validated through simulations and used to illuminate how waiting-time differences and service-rate heterogeneity drive switching behavior. The work contributes a decentralized perspective on queue management in network slices, accompanied by an open-source toolkit for simulating behavioral queuing in MEC/6G contexts, with implications for designing slice-aware resource sharing. Overall, it provides actionable insights into when jockeying reduces individual delays and how to balance local gains against system-wide performance in decentralized mobile networks.

Abstract

Use-case-specific network slicing in decentralized multi-tenancy cloud environments is a promising approach to bridge the gap between the demand and supply of resources in next-generation communication networks. Our findings associate different slice profiles to queues in a multi-server setting, such that tenants continuously assess their preferences and make rational decisions to minimize the queuing delay. Deviated from classical approaches that statistically model the jockeying phenomena in queuing systems, our work pioneers to setup a behavioral model of jockeying impatient tenants. This will serve as a basis for decentralized management of multi-queue systems, where the decision to jockey is individually made by each tenant upon its up-to-date assessment of expected waiting time. Additionally, we carry out numerical simulations to empirically unravel the parametric dependencies of the tenants' jockeying behavior.
Paper Structure (7 sections, 20 equations, 4 figures, 2 tables)

This paper contains 7 sections, 20 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Waiting time as determinant metric for jockeying decisions
  4. Expected jockeying frequency
  5. Assumptions during the experiment
  6. Empirical and Numerical Results
  7. Discussion and Future Work

Figures (4)

  • Figure 1: The distribution of the sizes of the buffers suggested a skewed orientation whose generalizations tended to Gaussian.
  • Figure 2: Sensitivity to variations in setup parameters: The figure on the left illustrated how combined variations in the processing rates of either queues affected the number of times the job was moved around. The graphic on the right in contrast captured the influence of these variations on the total time until service completion of the jockeyed job. It was observable that the differences in waiting times at the current position versus jockeyed to position had negligible effect on either measured descriptors.
  • Figure 3: A two dimensional extract that depicted the relationship between the difference in the capacities of the two buffers and the total time a jockeyed job consequently spent in the system against the jobs that were never migrated.
  • Figure 4: The jockeying customer consequently spent less time in the system. This profile augured with earlier studies to underline benefits of the behavior.