On the stability of a queueing system with uncountably branching fluid limits (Q2432953)

The paper considers a queueing system containing two stations and two servers. Station 1 receives an input stream of customers with rate \(\lambda _1<1\), Station 2 with rate \(\lambda _2>1\); both service rates are equal to 1. Having completed the service customers leave the system. The second server is always located at Station 2. The first server assists it from time to time - along with its own service at the first station. There are three possible states of the system. Neutral: each server works at its ``own'' station. The first is operating as an FCFS single server queue; the second station contains two buffers, the server serves the customers accumulated in one of them, while new customers are accumulated in the second one. When the first buffer becomes empty, they change place. State of readiness: the system operates as in neutral state, but the first customer is ready to come to the second station. The state of assistance: both servers operate at the second station taking customers from the same buffer. The system changes its states in the following order: neutral, readiness, assistance, then neutral again, etc. according to a given algorithm. There are two variants of stochastic assumptions: (SA1) the interarrival times of first stream are i.i.d. with mean \(1/\lambda _1\), for the second one Poisson with parameter \(\lambda _2\), the service times i.i.d. with mean 1; (SA2) the input streams are Poisson with parameters \(\lambda _1\) and \(\lambda _2\), the service times exponential with parameter 1. By using the fluid approximation one finds conditions for the stability and instability of the system and shows that fluid limits admit a branching with uncountable number of branches.