A Palm calculus approach to functional versions of Little's law (Q1805751)

Let \((T_n,W_n)\) be a queueing system with stationary arrival times \(T_n\) of intensity \(\lambda \) and service times \(W_n\) and suppose that for any \(k\leq n\), \(T_k+W_k\leq T_n+W_n\) \(P^0\)-almost surely, where \(P^0\) is the Palm distribution of the arrival process (FIFO assumption). For any time \(s\in {\mathbb R}\), let \(X(s)\) be the number of customers in the system at time \(s\) and let \(A(0,s]\) be the number of arrivals within the interval \((0,s]\). Using the Palm calculus, the author proves the following relation for the probability generating functions \[ E^0z^{X(0-)}=E^0z^{A(0,W_0]},\quad z\in {\mathbb C}. \] This is a generalization of the Little's law [\textit{S. Halfin} and \textit{W. Whitt}, Commun. Stat., Stochastic Models 5, No. 3, 515-529 (1989; Zbl 0677.60101)]. Assuming additionally that \(W_0\) and \(\{A(0,u]:u>0\}\) are \(P^0\)-independent (further arrivals do not affect the service times of customers already in the system), the author shows that \[ E^0z^{X(0-)}=\int_0^\infty (E^0z^{A(0,u]})F^0(du)\quad \text{and}\quad Ez^{X(0)}=1-\lambda (1-z)\int_0^\infty (1-F^0(u))(E^0z^{A(0,u]}) du, \] where \(F^0(u)=P^0(W_0<u)\).