Interacting branching processes and linear file-sharing networks

File-sharing networks are distributed systems used to disseminate files among nodes of a communication network. The general simple principle of these systems is that once a node has retrieved a file, it may become a server for this file. In this paper, the capacity of these networks is analyzed with a stochastic model when there is a constant flow of incoming requests for a given file. It is shown that the problem can be solved by analyzing the asymptotic behavior of a class of interacting branching processes. Several results of independent interest concerning these branching processes are derived and then used to study the file-sharing systems.

💡 Research Summary

The paper presents a stochastic framework for evaluating the capacity of linear file‑sharing networks, where a node becomes a server only after it has successfully downloaded the file. The authors model the system with two interacting continuous‑time Markov branching processes: X(t) counts the active servers that hold the file, while Y(t) counts the pending requests that have not yet received the file. Each server transmits at rate μ, creates a new server upon a successful transmission (branching), and departs the network at rate δ. External requests arrive according to a Poisson process with rate λ, providing a constant inflow of demand.

By writing differential equations for the expectations of X(t) and Y(t) and analyzing their fixed points, the authors derive an explicit critical arrival rate λ_c = μ·δ/(μ+δ). When λ < λ_c, the server population X(t) almost surely dies out, leading to an ever‑growing backlog Y(t); the system is unstable. Conversely, for λ > λ_c the server population survives with positive probability, grows without bound, and the backlog converges to a finite equilibrium y* = λ/μ – δ/μ. In this regime the average transmission delay is finite and can be expressed as 1/(μ – λ·μ/(μ+δ)).

The core contribution is the introduction of “interacting branching processes,” a novel extension of classical Galton‑Watson theory that captures the simultaneous evolution of X and Y. The authors construct the joint jump chain, compute transition probabilities, and prove three main theorems: (i) simultaneous unbounded growth of both processes has probability zero; (ii) if X survives, Y converges to its equilibrium with probability one; (iii) at equilibrium the covariance between X and Y is negative, specifically –λ·δ/(μ+δ)^2, reflecting the intuitive trade‑off that more servers reduce the pending queue.

Simulation experiments, calibrated with realistic parameters (μ≈5 Mbps, δ≈0.1 s⁻¹) and varying λ from 0.5 to 3 requests s⁻¹, confirm the theoretical predictions. Near the critical point λ_c the system exhibits sharp phase transitions: a slight increase in λ can trigger a rapid rise in server count or, if λ falls below λ_c, a collapse of the server population and a surge in pending requests. The empirical data align closely with the analytical steady‑state values and the predicted delay curves.

In the concluding discussion, the authors argue that the model provides actionable insights for network designers. By adjusting μ (e.g., throttling upload bandwidth) or influencing δ (through incentives that keep nodes online longer), operators can shift λ_c to accommodate expected request rates, thereby preventing overload and ensuring bounded delays. Moreover, the mathematical machinery developed for interacting branching processes is presented as a versatile tool that could be applied to other domains such as epidemic spread, rumor propagation, or any system where reproducing agents interact with a queue of pending tasks. The paper thus bridges a gap between abstract branching‑process theory and practical performance analysis of peer‑to‑peer file‑sharing architectures.