Fluid limits for interacting queues in sparse dynamic graphs
Goldsztajn, Borst, van Leeuwaarden
Consider a network of $n$ single-server queues where tasks arrive independently at each server at rate $λ_n$. The servers are connected by a graph that is resampled at rate $μ_n$ in a way that is symmetric with respect to the servers, and each task is dispatched to the shortest queue in the graph neighborhood where it appears. We aim to gain insight in the impact of the dynamic network structure on the load balancing dynamics in terms of the occupancy process which describes the empirical distribution of the number of tasks across the servers. This process evolves on the underlying dynamic graph, and its dynamics depend on the number of tasks at each individual server and the neighborhood structure of the graph. We establish that this dependency disappears in the limit as $n \to \infty$ when $λ_n / n \to λ$ and $μ_n \to \infty$, and prove that the limit of the occupancy process is given by a system of differential equations that depends solely on $λ$ and the limiting degree distribution of the graph. We further show that the stationary distribution of the occupancy process converges to an equilibrium of the differential equations, and derive properties of this equilibrium that reflect the impact of the degree distribution. Our focus is on truly sparse graphs where the maximum degree is uniformly bounded across $n$, which is natural in load balancing systems.
본 논문은 n개의 단일 서버 큐로 구성된 네트워크를 연구하며, 여기서 작업은 속도 λₙ으로 각 서버에 독립적으로 도착한다. 서버는 속도 μₙ으로 재샘플링되고 서버에 대해 대칭인 그래프를 통해 연결된다. 각 작업은 나타나는 그래프 이웃 중 가장 짧은 큐로 분배된다. 연구 목표는 동적 네트워크 구조가 부하 분산 동역학에 미치는 영향, 특히 점유 과정(서버 간 작업의 경험적 분포를 설명하는 과정)을 깊이 있게 이해하는 것이다. n→∞, λₙ/n→λ, μₙ→∞일 때, 이러한 의존성이 사라지며, 점유 과정의 극한은 λ와 그래프의 극한 차수 분포에만 의존하는 미분방정식 시스템으로 주어진다.