0% found this document useful (0 votes)
5 views2 pages

M/M/ M/M/ B: Queue With Impatient Customers

The document contains a series of problems related to applied probability and queueing theory, focusing on various types of queues such as M/M/∞, M/M/1, and G/M/1 systems. It explores topics like busy periods, invariant distributions, time-reversal properties, and the behavior of queues with impatient customers. Additionally, it discusses the unique solutions of traffic equations in Jackson networks and the implications for a specific insurance claim processing scenario.

Uploaded by

derekdereklch
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
5 views2 pages

M/M/ M/M/ B: Queue With Impatient Customers

The document contains a series of problems related to applied probability and queueing theory, focusing on various types of queues such as M/M/∞, M/M/1, and G/M/1 systems. It explores topics like busy periods, invariant distributions, time-reversal properties, and the behavior of queues with impatient customers. Additionally, it discusses the unique solutions of traffic equations in Jackson networks and the implications for a specific insurance claim processing scenario.

Uploaded by

derekdereklch
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Applied probability, Lent 2025. [email protected].

uk

Example Sheet 3

1. Compute the average busy period for a M/M/∞ and a M/M/1 queue. (The busy period
B is the length of time between the arrival of the first customer and the first time afterwards
that all servers are free).
2. Consider the M/M/n queue, where the arrival rate is λ and the service rate in each queue
is µ. For which values of the parameters is the queue length transient, positive recurrent and
null recurrent? Compute the invariant distribution when there exists one.
3. Queue with impatient customers. Customers arrive at a single server at rate λ and require
an exponential amount of service with rate µ. Customers waiting in line are impatient and
if they are not in service they will leave at rate δ, independent of their position in the queue.
(a) Show that for any δ > 0 the system has an invariant distribution. (b) Find the invariant
distribution when δ = µ.
4. Is the time-reversal of a tandem of M/M/1 queues reversible at equilibrium? If not can
you describe the time-reversal? Find the distribution of the process Dt which counts the
number of customers whose service is completed by time t.
5. Consider the following queue. Customers arrive at rate λ > 0 and are served by one server
at rate µ. After service, each customer returns to the beginning of the queue with probability
p ∈ (0, 1). Let (Lt )t≥0 denote the queue length. Show that L has the same distribution as a
M/M/1 queue with modified rates. For which parameters is L transient, and for which is it
recurrent?
6. Let (Lt )t≥0 denotes the length of an M/M/1 queue with rates λ < µ. Let π denote
the equilibrium distribution. Let (Dt )t≥0 denotes the departure process from the queue. By
considering all possibilities leading to the events below, show directly that as h → 0,

Pπ (Dh − D0 = 0) = 1 − λh + o(h)

and that
Pπ (Dh − D0 ≥ 1) = λh + o(h).
What have you proved?
7. Prove that the traffic equations for a Jackson network have a unique solution.
8. Let Xt = (Xt1 , . . . , XtN , t ≥ 0) denote a Jackson network of N queues, with arrival rate λi
and service rate µi in queue i, and eachP customer moves to queue j ̸= i with probability pij
after service from queue i. We assume j pij < 1 for each i = 1, . . . , N and that the traffic
equations have a solution such that λ̄i < µi .
Describe the time-reversal of X at equilibrium.
Let Di (t) be the process of (final) departures from queue i. Show that, at equilibrium,
(Di (t), t ≥ 0)1≤i≤N are independent Poisson processes and specify the rates. Show further
that Xt is independent (Di (s), 1 ≤ i ≤ N, 0 ≤ s ≤ t).
9. Consider a system of N queues serving a finite number K of customers. The system
evolves as follows. At station 1 ≤ i ≤ N , customers are served one at a time at rate µi . After

1
service, each customer P moves to queue j with probability pij > 0. We assume here that the
system is closed, ie, j pij = 1 for all 1 ≤ i ≤ N .
Let S = {(n1 , . . . , nN ) : ni ∈ N, N
P
i=1 ni = K} be the state space of the Markov chain.
Write down its Q-matrix. Also write down the Q-matrix R corresponding to the position in
the network of one customer (that is, when K = 1). Show that there is a unique distribution
(λi )1≤i≤n such that λR = 0. Show that
N
Y
π(n) = CN λni i , n ∈ S
i=1

defines an invariant measure for the chain. Are the queue lengths independent at equilibrium?
10. Kafkaian Insurances Inc. has a peculiar way of processing claims. Claims arrive at a rate
of 10 per day, and are initially randomly assigned to one of two departments, respectively D1
and D2 . The service rates in D1 and D2 are µ1 = 15 and µ2 = 20 per day, respectively. After
looking at each claim, the relevant department settles the claim with probability 1/2, and
otherwise finds a pretext to hand it over to the other department to process it. This goes on
until the claim is finally settled by one of the two departments.
(a) What proportion of claims is finally settled by D1 ?
(b) How many claims are settled on average every month by Kafkaian Insurances Inc.?
(c) The manager of the company wants to reward the work of his employees based on the
number of claims that their department settles. Is that a good idea?
11. Consider a G/M/1 queueing system: the nth client arrives at time An = ni=1 ξi , where
P
(ξi ) is a sequence of nonnegative i.i.d. random variables, and the service times are i.i.d.
exponential with rate µ. Let Xn = L(An ) be the size of the queue just before the nth arrival.
(i) Show that (Xn ) is a discrete-time Markov chain, and specify its transition matrix.
(ii) Show that if ρ := (µEA)−1 < 1 then the chain (Xn ) has a unique equilibrium distri-
bution π = (πi ) and hence is positive recurrent. Here
πi = (1 − η)η i , i = 0, 1, . . .
and η ∈ (0, 1) is a solution to η = ϕ(µ(η − 1)), where for θ ∈ R, ϕ(θ) = E(eθξ ).

12. Consider the square lattice Z2 , and endow each site x ∈ Z2 with a weight Wx , which
is an independent exponential random variable of rate µ. An oriented path π between (1, 1)
and a point (M, N ), with M, N ≥ 1, is called increasing if it only ever goesPin the North and
East directions. Define the weight of an increasing path π to be W (π) = x∈π Wx , and the
passage time from (1, 1) to (M, N ) to be
T (M, N ) = max W (π)
π

where the max is over increasing π’s from (1, 1) to (M, N ). This model is called Last Passage
Percolation. [Simulations showing optimal paths from (0,0) are interesting.]
The goal of this question is to relate this model to a sequence of N queues operating under
the following protocol. At time 0 there are M customers in the first queue, and none at any
other queue. Customers are served one at a time at rate µ in each queue, and after service
at queue i, a customer moves on to queue i + 1. Customers leave the system for good after
being served at queue N . Let τ (M, N ) denote the time at which the M th customer completes
service in queue N . Show that τ (M, N ) and T (M, N ) have the same distribution.

You might also like