B4.12

Applied Probability
Part II, 2004

Consider an M/G/1M / G / 1 queue with ρ=λES<1\rho=\lambda \mathbb{E} S<1. Here λ\lambda is the arrival rate and ES\mathbb{E} S is the mean service time. Prove that in equilibrium, the customer's waiting time WW has the moment-generating function MW(t)=EetWM_{W}(t)=\mathbb{E} e^{t W} given by

MW(t)=(1ρ)tt+λ(1MS(t))M_{W}(t)=\frac{(1-\rho) t}{t+\lambda\left(1-M_{S}(t)\right)}

where MS(t)=EetSM_{S}(t)=\mathbb{E} e^{t S} is the moment-generating function of service time SS.

[You may assume that in equilibrium, the M/G/1M / G / 1 queue size XX at the time immediately after the customer's departure has the probability generating function

EzX=(1ρ)(1z)MS(λ(z1))MS(λ(z1))z,0z<1.]\left.\mathbb{E} z^{X}=\frac{(1-\rho)(1-z) M_{S}(\lambda(z-1))}{M_{S}(\lambda(z-1))-z}, \quad 0 \leqslant z<1 .\right]

Deduce that when the service times are exponential of rate μ\mu then

MW(t)=1ρ+λ(1ρ)μλt,<t<μλ.M_{W}(t)=1-\rho+\frac{\lambda(1-\rho)}{\mu-\lambda-t}, \quad-\infty<t<\mu-\lambda .

Further, deduce that WW takes value 0 with probability 1ρ1-\rho and that

P(W>xW>0)=e(μλ)x,x>0.\mathbb{P}(W>x \mid W>0)=e^{-(\mu-\lambda) x}, \quad x>0 .

Sketch the graph of P(W>x)\mathbb{P}(W>x) as a function of xx.

Now consider the M/G/1M / G / 1 queue in the heavy traffic approximation, when the service-time distribution is kept fixed and the arrival rate λ1/ES\lambda \rightarrow 1 / \mathbb{E} S, so that ρ1\rho \rightarrow 1. Assuming that the second moment ES2<\mathbb{E} S^{2}<\infty, check that the limiting distribution of the re-scaled waiting time W~λ=(1λES)W\tilde{W}_{\lambda}=(1-\lambda \mathbb{E} S) W is exponential, with rate 2ES/ES22 \mathbb{E} S / \mathbb{E} S^{2}.