Interference Queueing Networks: A Replica Mean-Field Approach in the Symmetric Setting

arXiv:2606.13264 · math.PR, math-ph, math.MP · Submitted 2026-06-11 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Mathematics. The day in mathematics: new theorems, new proofs, and what they open.

Ada: We're looking at the paper: "Interference Queueing Networks".

Emmy: Here's the gist of the paper. We propose a model for evaluating the performance of wireless communication networks beyond the ubiquitous full-buffer assumption, under which every transmitter is always active This paper introduces a tractable model for analyzing coupled queueing systems where service rates depend on neighboring interference, providing guidance for network dimensioning and engineering Model Setup and Assumptions The network is represented by N interacting queues arranged on a torus, with homogeneous arrival rate λ and service rates depending on the activity of neighboring interferers Each queue is associated with a transmitter–receiver pair, and its service rate is given by the Shannon capacity, which depends on the corresponding Signal-to-Interference-plusNoise Ratio (SINR) The service process is therefore state-dependent, and all queues become coupled through interference Key assumptions include one queue per wireless link, short links where receiver–transmitter distances are small compared with intertransmitter distances, homogeneous queues with common arrival rate and file-size distribution, and symmetric geometry on a torus Stability Analysis and Thresholds The first contribution is to make the intuition rigorous by proving the existence of a critical threshold for the arrival rate The system is stable, in the sense that it returns to the empty state infinitely often, if λ < µ∗ This threshold has a natural interpretation: the arrival rate must be smaller than the worst-case transmission rate µ∗, corresponding to the rate at which a queue transmits when all surrounding queues are active The stability region is characterized by Theorem 3.1, stating that for λ > µ∗, the process Q is transient, whereas for λ < µ∗, it is positive recurrent Mean-Field Approximations To investigate the stationary regime, two mean-field approximations are introduced The first approximation is obtained from a finite K-replica system, for which propagation of chaos is proven as K → ∞ The second approximation describes a queueing system evolving in a suitably chosen autonomous environment These approximations lead to different classes of tractable models, namely queues in a random autonomous environment and McKean–Vlasov systems Performance Comparison and Convergence The two mean-field approximations capture different aspects of the system: the first approximation performs well for the mean delay and congestion, whereas the second approximation is more accurate for the busy probability The conjecture states that the stationary busy probability of the original system lies between those predicted by these two mean-field models Furthermore, a Foster–Lyapunov drift condition is established to prove exponential ergodicity of the true system Long-Term Behavior and Randomness The long-time behavior of the true system converges exponentially fast to stationarity The analysis extends to systems with randomly spaced queues, where common spacings are taken uniformly distributed on

a, b: This extension shows that the mean-field results can still be applied in this more general spatial setting The analysis also provides quantitative estimates for the expected delay of typical job using the Mean-Field 1 approximation Open Problems and Extensions Several open problems remain, including whether one can derive quantitative error bounds between the true system and the proposed mean-field approximations Can uniqueness be proved for the fixed-point equation (9) The paper also suggests future work on relaxing assumptions such as equally spaced transmitters and incorporating feedback mechanisms to improve performance Appendix The appendix contains the technical results, including Lemma A.1 showing there exist a path wise unique solution to (3) It also includes Lemma A.3 establishing the monotonicity of the Markov chain Q with respect to the arrival rate, initial condition, and departure rate Lemma A.1 There exist a path wise unique solution to (3) Proposition A.2 If Q and Q˜ have the same law for the initial condition then Q d= Q˜ Lemma A.3 The Markov chain Q is monotone with respect to the arrival rate, the initial condition, and the departure rate Theorem 3.12 The McKean–Vlasov equation (7) admits a strong solution which is pathwise unique Lemma 4.5 The McKean–Vlasov equation (7) admits a unique strong solution on

0, Te: , for every T >e 0 such that T Lξ e N < 1 Theorem 3.14 This converges to 0 as K → ∞ Lemma A.

Paper summary: Ada: So this paper is called "Interference Queueing Networks: A Replica Mean-Field Approach in the Symmetric Setting," and it looks at wireless networks when you go beyond that simple idea that every transmitter is always on.

Emmy: Exactly. The authors propose a model where we have N interacting queues arranged on a torus, and the service rate of each queue actually depends on how busy its neighbors are because of interference.

Ada: They set up this system with a homogeneous arrival rate lambda, and the service rate is based on the Shannon capacity, which means it depends directly on the Signal-to-Interference-plus-Noise Ratio, or SINR.

Kolya: So the core idea is that when neighbors are empty, your service rate gets better because there's less interference hitting you. That's a key mechanism they are modeling here.

Emmy: They make several assumptions to keep things tractable: they assume short links where receiver and transmitter distances are small compared to the distance between transmitters, and they have homogeneous queues with the same arrival rate and file size distribution.

Ada: The paper then goes on to prove something really important: they establish a critical threshold for that arrival rate lambda. Stability means the system comes back to an empty state infinitely often if lambda is smaller than this specific value, mu star.

Kolya: That threshold mu star has a physical meaning: it's the maximum transmission rate you can have before the entire system becomes unstable, which happens when every single queue is transmitting at its worst case rate because all neighbors are active.

Emmy: And to figure out what happens in the steady state, they introduce two different mean-field approximations. The first one comes from a finite K-replica system, and they prove that propagation of chaos holds as K goes to infinity.

Ada: That second approximation describes the system evolving in a specific kind of random environment, leading to models like queues in a random autonomous environment or McKean–Vlasov systems.

Kolya: Those two approximations are used for different things; the first one is better for understanding the mean delay and congestion, while the second one is more accurate when you're looking at the busy probability of being busy.

Emmy: The paper suggests that if you look at both these mean-field models, they predict a range for what we call the stationary busy probability of the original system.

Conclusion: Ada: So, looking at "Interference Queueing Networks: A Replica Mean-Field Approach in the Symmetric Setting," the authors tackle how interference affects queueing networks in wireless communication settings where transmitters are always active.

Emmy: It’s about building a tractable model for coupled queueing systems where service rates change based on neighboring interference, specifically using a replica mean-field approach.

Ada: The main claim is that they rigorously prove the existence of a critical threshold for the arrival rate lambda, which tells us when the system is stable and when it's transient.

Kolya: What this means practically is that you need to keep your traffic rate below this threshold mu star for the network to stay in a steady state, because if you go above it, things just explode.

Emmy: The paper uses those mean-field approximations to get different insights into the system's long-term behavior and stationary distribution.

Ada: It really connects theoretical queueing theory with the specific physics of interference in these geometric network layouts.

Kolya: The extension to randomly spaced queues also shows that these mean-field results still hold even when the transmitters aren't perfectly equally spaced on the torus, though they do have open questions about deriving quantitative error bounds between those approximations and the true system.

inria

math.PR, math-ph, math.MP

Submitted: 2026-06-11

Updated: 2026-11-04

Code: https://github.com/Philippe-Sarotte/Interference-Queueing-Networks-A-ReplicaMean-Field-Approach-in-the-Symmetric-Setting

License: http://creativecommons.org/licenses/by-nc-sa/4.0/

Importance score: 91/100

The gist: We propose a model for evaluating the performance of wireless communication networks beyond the ubiquitous full-buffer assumption…

Key concepts

Coupled Queueing Systems
This refers to a network of interacting queues where the service rate of one queue depends on the activity (interference) of its neighbors. In this wireless context, a transmitter's speed is limited by how much signal it receives from other active transmitters nearby.
Mean-Field Approximations
These are mathematical techniques used to simplify complex systems by assuming that the behavior of one individual element can be approximated by the average behavior of the entire system. The paper uses two such approximations—one based on a finite replica system and another describing evolution in an autonomous environment—to create simpler, solvable models.
Critical Threshold ($μ^*$)
This is a specific arrival rate limit that determines whether the network is stable or unstable. If the arrival rate ($μ$) is below this threshold, the system returns to an empty state infinitely often (stable). If $μ$ exceeds it, the process becomes transient (unstable). This threshold represents the maximum sustainable transmission rate.
McKean–Vlasov Systems
This is a class of mathematical models that describe how a system evolves when its individual components interact with each other's distribution. In this paper, it is one of the tractable models derived from the mean-field approximations, used to study the stationary regime of the interference queueing network.

Terminology

Summary

We propose a model for evaluating the performance of wireless communication networks beyond the ubiquitous full-buffer assumption, under which every transmitter is always active This paper introduces a tractable model for analyzing coupled queueing systems where service rates depend on neighboring interference, providing guidance for network dimensioning and engineering

Model Setup and Assumptions

The network is represented by N interacting queues arranged on a torus, with homogeneous arrival rate λ and service rates depending on the activity of neighboring interferers Each queue is associated with a transmitter–receiver pair, and its service rate is given by the Shannon capacity, which depends on the corresponding Signal-to-Interference-plusNoise Ratio (SINR) The service process is therefore state-dependent, and all queues become coupled through interference Key assumptions include one queue per wireless link, short links where receiver–transmitter distances are small compared with intertransmitter distances, homogeneous queues with common arrival rate and file-size distribution, and symmetric geometry on a torus

Stability Analysis and Thresholds

The first contribution is to make the intuition rigorous by proving the existence of a critical threshold for the arrival rate The system is stable, in the sense that it returns to the empty state infinitely often, if λ < µ∗ This threshold has a natural interpretation: the arrival rate must be smaller than the worst-case transmission rate µ∗, corresponding to the rate at which a queue transmits when all surrounding queues are active The stability region is characterized by Theorem 3.1, stating that for λ > µ∗, the process Q is transient, whereas for λ < µ∗, it is positive recurrent

Mean-Field Approximations

To investigate the stationary regime, two mean-field approximations are introduced The first approximation is obtained from a finite K-replica system, for which propagation of chaos is proven as K → ∞ The second approximation describes a queueing system evolving in a suitably chosen autonomous environment These approximations lead to different classes of tractable models, namely queues in a random autonomous environment and McKean–Vlasov systems

Performance Comparison and Convergence

The two mean-field approximations capture different aspects of the system: the first approximation performs well for the mean delay and congestion, whereas the second approximation is more accurate for the busy probability The conjecture states that the stationary busy probability of the original system lies between those predicted by these two mean-field models Furthermore, a Foster–Lyapunov drift condition is established to prove exponential ergodicity of the true system

Long-Term Behavior and Randomness

The long-time behavior of the true system converges exponentially fast to stationarity The analysis extends to systems with randomly spaced queues, where common spacings are taken uniformly distributed on [a, b] This extension shows that the mean-field results can still be applied in this more general spatial setting The analysis also provides quantitative estimates for the expected delay of typical job using the Mean-Field 1 approximation

Open Problems and Extensions

Several open problems remain, including whether one can derive quantitative error bounds between the true system and the proposed mean-field approximations Can uniqueness be proved for the fixed-point equation (9) The paper also suggests future work on relaxing assumptions such as equally spaced transmitters and incorporating feedback mechanisms to improve performance

Appendix

The appendix contains the technical results, including Lemma A.1 showing there exist a path wise unique solution to (3) It also includes Lemma A. p. 9

Where the paper says it

12 (Well-posedeness of McKean–Vlasov). (7) admits a strong solution which is path-wise unique.

3 establishing the monotonicity of the Markov chain Q with respect to the arrival rate, initial condition, and departure rate

Lemma A.1

There exist a path wise unique solution to (3)

Proposition A.2

If Q and Q˜ have the same law for the initial condition then Q d= Q˜

Lemma A.3

The Markov chain Q is monotone with respect to the arrival rate, the initial condition, and the departure rate

Theorem 3.12

The McKean–Vlasov equation (7) admits a strong solution which is pathwise unique

Lemma 4.5

The McKean–Vlasov equation (7) admits a unique strong solution on [0, Te], for every T >e 0 such that T Lξ e N < 1

Theorem 3.14

This converges to 0 as K → ∞

Lemma A.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Replica mean-field approximation for performance analysis allows AI systems to estimate mean delay and congestion levels by replacing random service rates with averaged counterparts, leading to tractable M/M/1 type models for computation.

  2. McKean–Vlasov approximation of type 1 enables the derivation of a stationary distribution given by a product measure where marginals are solved from the fixed-point equation λ x = B log2 (1 + 1 / (1 + x ξN)).

  3. The second mean-field approximation, QAMF2, provides an estimate for the busy probability that is less congested than Mean-Field 1, suggesting a better predictive tool for congestion level estimation in certain regimes.

  4. Exponential convergence of the true system to stationarity allows AI systems to obtain stationary samples without simulating excessively long time horizons by using a Foster–Lyapunov drift condition where LV <= -cV + b1C.

  5. The analysis of random positions provides formulas for the expected number of jobs E[q] using the Mean-field approximation EMF1[q], which can be computed in terms of the common distance S between queues.

Abstract

We propose a model for evaluating the performance of wireless communication networks beyond the ubiquitous full-buffer assumption, under which every transmitter is always active. The network is represented by N interacting queues arranged on a torus, with homogeneous arrival rate and service rates depending on the activity of neighboring interferers. More precisely, each queue is associated with a transmitter-receiver pair, and its service rate is given by the Shannon capacity, which depends on the corresponding Signal-to-Interference-plus-Noise Ratio (SINR). Since interfering transmitters only emit when their queue is non-empty, the SINR, and hence the service rate, improves when neighboring queues are empty. We first derive the stability region of the system. To investigate the stationary regime, we introduce two mean-field approximations. The first one is obtained from a finite K-replica system, for which we prove propagation of chaos as K goes to infinity. The second one describes a queueing system evolving in a suitably chosen environment. We derive quantitative estimates comparing these two approximations. Finally, we prove that the original interacting queueing system converges exponentially fast to stationarity.

Sources

Related papers