Large-time asymptotics for the \(G_{t}/M_{t}/s_{t}+GI_{t}\) many-server fluid queue with abandonment (Q632220)

The \(G_{t}/M_{t}/s_{t}+GI_{t}\) many-server fluid queue with time-varying parameters is approximated. An asymptotic loss of memory (ALOM) property for that fluid model is established. It is shown that, for this model there exists an asymptototic independence from the initial conditions as time \(t\) evolves under regularity conditions. Moreover, the difference in the performance functions dissipates over exponentially fast time, again under regular conditions. The transient performance of the stationary \(G/M/s + GI\) fluid queue converges with its steady state performance, i.e., the periodic \(G_{t}/M_{t}/s_{t}+GI\) fluid queue approaches the periodic steady state as time evolves for all finite initial conditions. This is the main result of the paper. The special case of a periodic fluid model is also studied. The conditions under which there exists a uniqueue periodic steady state (PSS) for a periodic fluid model and the time-varying performance to that PSS for all finite initial conditions are formulated. Finally, some numerical results of the periodic steady state of the \(G_{t}/M_{t}/s_{t} + M\) are compared with simulation results. This paper exploits deterministic fluid models in order to approximately solve important control problems for stochastic systems.