Applied Statistics & Queuing Theory

Course No: CSE 3107

Topic : Queuing Theory

Julia Rahman
 Queuing Theory
 A queue = a line of people or things waiting to be served
 Queueing Theory – ways of predicting how long the line
or the wait will be, or deciding how many servers to have
 Some people spell it queuing rather than queueing;
 Mathematical analysis of queues and waiting times in
stochastic systems.
 Used extensively to analyze production and service processes
exhibiting random variability in market demand (arrival times) and
service times.
 Queues arise when the short term demand for service
exceeds the capacity
 Most often caused by random variation in service times and the
times between customer arrivals.
 If long term demand for service > capacity the queue will explode!
 Queuing Theory
Why is Queuing Analysis Important?
 Capacity problems are very common in industry and one
of the main drivers of process redesign
 Need to balance the cost of increased capacity against the gains
of increased productivity and service
 Queuing and waiting time analysis is particularly
important in service systems
 Large costs of waiting and of lost sales due to waiting
 Prototype Example – ER at County Hospital
 Patients arrive by ambulance or by their own accord
 One doctor is always on duty
 More and more patients seeks help → longer waiting times
 Question: Should another MD position be instated?
 Queuing Theory

Total
cost
Cost

Cost of
service

Cost of waiting

Process capacity
 Queuing Theory
 The subject of queueing theory can be described as
follows:
 A service center and a population of customers, which at
some times enter the service center in order to obtain
service.
 Service center can only serve a limited number of
customers.
 If a new customer arrives and the service is exhausted, he
enters a waiting line and waits until the service facility
becomes available.
 Three main elements of a service center:
1. a population of customers,
2. the service facility and
3. the waiting line.
 Queuing Theory
Applications:
 Commercial Queuing Systems
 Commercial organizations serving external customers
 Ex. Dentist, bank, ATM, gas stations, plumber, garage …
 Transportation service systems
 Vehicles are customers or servers
 Ex. Vehicles waiting at toll stations and traffic lights, trucks or
ships waiting to be loaded, taxi cabs, fire engines, buses
 Business-internal service systems
 Customers receiving service are internal to the organization
providing the service
 Ex. Inspection stations, conveyor belts, computer support …
 Social service systems
 Ex. Judicial process, the ER at a hospital, waiting lists for
organ transplants or student dorm rooms …
 Application of queuing models in
Computer Science
 Queueing theory in networking:
 Dimensioning of buffers in routers or multiplexers
 Number of trunks in a central office in POTS,
 Calculating end-to-end throughput in networks and so forth.
 Server virtualized system with live VM migration
 Service delivery improvements for cloud service
providers
 Trading power consumption against performance by
reserving blocks of servers
 Optimal partitioning of a multi-core server processor
 Modeling and optimizing the delay-energy tradeoff in
TDM systems with sleep mode
 Queuing System
 model processes in which customers arrive.
 wait their turn for service.
 are serviced and then leave.
 Example: supermarket checkouts stands, world series
ticket booths, doctors waiting rooms etc..
 Five components of a Queueing system:
 Interarrival-time probability density function (pdf)
 Service-time
 Number of servers
 Queueing discipline
 Size of queue.
 Queuing Theory
Components of a Basic Queuing Process :
Input Source The Queuing System

Served
Calling Jobs Service Jobs
Queue Mechanism
Population
leave the
system

Arrival Queue
Process Discipline
Service
Queue
Process
Configuration
 Queuing System
 Assume: an infinite number of customers (i.e. long
queue does not reduce customer number).
 Assumption is bad in :
 a time-sharing model.
 with finite number of customers.
 if half wait for response, input rate will be reduced.
 Queuing System
The calling population
 The population from which customers/jobs originate
 The size can be finite or infinite (the latter is most common)
 Can be homogeneous (only one type of customers/ jobs) or
heterogeneous (several different kinds of customers/jobs)
The Arrival Process
 Determines how, when and where customer/jobs arrive to
the system
 Important characteristic is the customers’/jobs’ inter-arrival
times
 To correctly specify the arrival process requires data
collection of interarrival times and statistical analysis
 Queuing System
The queue configuration
 Specifies the number of queues
 Single or multiple lines to a number of service stations
 Their location
 Their effect on customer behavior
 Balking and reneging
 Their maximum size (number of jobs the queue can
hold)
 Distinction between infinite and finite capacity
 Queuing System
The queue configuration (cont.)

Multiple Queues Single Queue

Servers Servers
 Queuing System
The queue configuration (cont.)
Multiple Line Advantages Single Line Advantages
1. The service provided can be 1. Guarantees fairness
differentiated  FIFO applied to all arrivals
 Ex. Supermarket express 2. No customer anxiety
lanes regarding choice of queue
2. Labor specialization possible 3. Avoids “cutting in” problems
3. Customer has more flexibility 4. The most efficient set up for
4. Balking behavior may be minimizing time in the queue
deterred 5. Jockeying (line switching) is
 Several medium-length lines are avoided
less intimidating than one very
long line
 Queuing System
The Service Mechanism
 Can involve one or several service facilities with one or
several parallel service channels (servers) -
Specification is required
 The service provided by a server is characterized by its
service time
 Specification is required and typically involves data
gathering and statistical analysis.
 Most analytical queuing models are based on the
assumption of exponentially distributed service times,
with some generalizations.
 Queuing System
The queue discipline
 Specifies the order by which jobs in the queue are
being served.
 Most commonly used principle is FIFO.
 Other rules are, for example, LIFO, SPT, EDD…
 Can entail prioritization based on customer type.
 Queuing System
Mitigating Effects of Long Queues
1. Concealing the queue from arriving customers
 Ex. Restaurants divert people to the bar or use pagers,
amusement parks require people to buy tickets outside
the park, banks broadcast news on TV at various
stations along the queue, casinos snake night club
queues through slot machine areas.
2. Use the customer as a resource
 Ex. Patient filling out medical history form while waiting
for physician
3. Making the customer’s wait comfortable and distracting
their attention
 Ex. Complementary drinks at restaurants, computer
games, internet stations, food courts, shops, etc. at
airports
 Queuing System
Mitigating Effects of Long Queues (cont.)
4. Explain reason for the wait
5. Provide pessimistic estimates of the remaining wait
time
 Wait seems shorter if a time estimate is given.
6. Be fair and open about the queuing disciplines used
 Stochastic Processes
 Stochastic process: System that changes over time in
an uncertain manner.
A collection of random variables {Xt}, where t  T = {0, 1, 2, .
. .}.
 State: Snapshot of the system at some fixed point in
time
 Transition: Movement from one state to another

a a a a

0 1 2 3

d d d d
 Stochastic Processes
 Definition: A stochastic process in continuous time is a
family {X(t)} of stochastic variables defined over a
continuous set of t-values.
 Example: The number of phone calls connected through
a switch board

X(t)=# Calls

 Definition: A stochastic process {X(t)} is said to have

independent increments if for all disjoint intervals (ti, ti+hi)
the differences Xi(ti+hi)Xi(ti) are mutually independent.
 Stochastic Processes

 Stochastic Processes
 Discrete time, Discrete state space: number of
accidents at 9 pm, 10 pm, 11 pm etc.
 Continuous time, Discrete state space: Number of
accidents on a highway at an interval of time say 9 pm to
10 pm.
 Discrete time, continuous state space: age of people
in a city at a particular year say 2001, 2002, 2003 etc.
 Continuous time, Continuous state space: age of
people from March 2013 to march 2014.
 Markovian Property
 Given that the present state is known, the conditional
probability of the next state is independent of the states
prior to the present state.
 Present state at time t is i: Xt = i
 Next state at time t + 1 is j: Xt+1 = j

 Conditional Probability Statement of Markovian Property:

Pr{Xt+1 = j | X0 = k0, X1 = k1,…,Xt = i} = Pr{Xt+1 = j | Xt = i}

for t = 0, 1,…, and all possible sequences i, j, k0, k1, . . . , kt–1.

 Markov Chain
 A discrete state space
 Markovian property for transitions
 One-step transition probabilities, pij, remain constant
over time (stationary)
 A discrete parameter Markov process is known as
Markov chain
 The following state-transition diagram is called a Markov
chain model.
 Directed branches represent transitions between the
states.
Po  P1  P k -1 P k
0 1 2 ... k-1 k k+1

 P1 P 2 P k  P k +1
 Markov Chain
not (4,7) not (5,7) not (6,7) not (8,7) not (9,7) not (10,7)

P4 P5 P6 P8 P9 P10

5 6 8
4 9 7 7 7
7
5 6 8 9 7
Win 10 7
4 10 Lose
(7, 11) Start (2, 3, 12)

State-Transition Network for Craps

 Markov Chain
 State-Transition Network for Simple Markov Chain

(0.6)

1
(0.1) 1 2 3
(0.3)
1 0.6 0.3 0.1
(1) (0.8)
 
P= 2  0.8 0.2 0 
3 1 
3 2
0 0
(0.2)
 Markov Chain
Classification of states:
 Accessible: Possible to go from state i to state j (path exists
in the network from i to j).
 Two states communicate if both are accessible from each
other. A system is irreducible if all states communicate.
 State i is recurrent if the system will return to it after leaving
some time in the future.
 If a state is not recurrent, it is transient.
d1 d2 d3 d4
0 1 2 3 4 …
a0 a1 a2 a3

0
a0
1
a1
2
a2
3
a3
4 …
 Markov Chain
Classification of states (cont.):
 A state is periodic if it can only return to itself after a fixed
number of transitions greater than 1 (or multiple of a fixed
number).
 A state that is not periodic is aperiodic. A state with a self-
loop transition (i.e. pii > 0) is always aperiodic.
(0.5)
 0 0 4
(0.5)
(1) (1) (1) (1)

2 1 2 1
(1) (1)

a. Each state visited b. Each state visited in multiples

every 3 iterations of 3 iterations
 Markov Chain


d1 d2 d3
0 1 2 3 4
a1 a2 a3
 Markov Chain
Classification of states (cont.):
 Class: set of states that communicate with each other.
 A class is either all recurrent or all transient and may be either
all periodic or aperiodic.
 States in a transient class communicate only with each other
so no arcs enter any of the corresponding nodes in the
network diagram from outside the class. Arcs may leave,
though, passing from a node in the class to one outside.

0 3 6

2 1 5 4
 Markov Chain
Illustration of Concepts:
 Example 1
0

State 0 1 2 3
0 0 X X 0
1 X 0 0 0 3 1
2 0 0 0 X
3 X 0 0 X

 Every pair of states communicates, forming a single recurrent

class; however, the states are not periodic.
 Thus the stochastic process is aperiodic and irreducible.
 Markov Chain
Illustration of Concepts:
 Example 2

State 0 1 2 3 4
0 X X 0 0 0
1 X X 0 0 0 4 1
2 0 0 X 0 0
3 0 0 X X 0
4 X 0 0 0 0

3 2

 States 0 and 1 communicate and for a recurrent class.

 States 3 and 4 form separate transient classes.
 State 2 is an absorbing state and forms a recurrent class.
 Markov Chain
Illustration of Concepts:
 Example 3
0

State 0 1 2 3
0 0 X X 0
1 0 0 0 X 1
3
2 0 0 0 X
3 X 0 0 0

 Every state communicates with every other state, so we have

irreducible stochastic process.
 Periodic? Yes, so Markov chain is irreducible and periodic.
 Discrete Time Markov Chain
 Definition: a stochastic process with discrete state
space and discrete time {Xn, n > 0} is a discrete time
Markov Chain (DTMC) iff
P[Xn+1 = j  Xn = in, ..., X0 = i0] = P[Xn+1 = j  Xn = in] = pij(n)
 In a DTMC, the past history impacts on the future
evolution of the system via the current state of the
system
 pij(n) is called transition probability from state i to
state j at time n.
 Discrete Time Markov Chain
Stochastic
process

Discrete Continuous
events event

Discrete Continuous A DTMC is a discrete

time time
time and memoriless
discrete event
stochastic process.
Memoryless
 Discrete Time Markov Chain
 Example: a mouse in a maze

1 2 start

5 3 4

exit 0

 Which stochastic process can be used to represent the

position of the mouse at time t?
 Under which assumptions, the system can be
represented by a discrete time Markov chain?
 Discrete Time Markov Chain
 Example: a mouse in a maze
1/2 1/2
1 2 start
1/2
1/4 1/2
1/4
1
1/4
5 3 4
1
1/4

exit 0
1
 Let {Xn}n=0, 1, 2, ... the position of the mouse after n rooms
visited
 Assume that the mouse does not have any memory of
rooms visited previously and that she chooses any corridor
equiprobably.
 Discrete Time Markov Chain
Homogenuous DTMC
 A DTMC is said homogenuous iff its transitions
probabilities do not depend on the time n, i.e.
P[Xn+1 = j  Xn = i] = P[X1 = j  X0 = i] = pij
 A homogenuous DTMC is then defined by its transition
matrix P =[pij]i,jE
 Discrete Time Markov Chain
Stochastic Matrix
 A square matrix is said stochastic iff
 all entries are non negative
 each line sums to 1
 Properties
• A transition matrix is a stochastic matrix
• If P is stochastic, then Pn is stochastic
• The eigenvalues of P are all smaller than 1, i.e. |l| ≤1
 Discrete Time Markov Chain
 A DTMC is said irreducible iff a state j can be reached
in a finite number of steps from any other state i.

 An irreducible DTMC is a strongly connected graph.

 Continuous Time Markov Chain
Stochastic
process

Discrete Continuous
events event

Discrete Continuous A CTMC is a

time time
continuous time and
memoriless discrete
event stochastic
Memoryless process.
Continuous Time Markov Chain
 Definition : a stochastic process with discrete state
space and continuous time {X(t), t > 0} is a continuous
time Markov Chain (CTMC) iff
P[X(t+s)= j X(u), 0≤u≤s] = P[X(t+s)= j X(s)], "t, "s, "j

Memoryless:
 In a CTMC, the past history impacts on the future
evolution of the system via the current state of the
system
Continuous Time Markov Chain

Poisson Exponential
Arrivals service time

N(t) : number of
customers at time t

Customer Customer
Arrivals departures
Continuous Time Markov Chain
 Definition : A CTMC {X(t), t > 0} is homogeneous iff
P[X(t+s)= j X(t) = i] = P[X(t+s)= j X(t) = i] = pij(s)

 Homogeneous memoryless:
In reliability, we only say "a machine that does not fail at
age t is as good as new"

 Only homogeneous CTMC will be considered in

this chapter.
 Continuous Time Markov Chain
Behaviour of a CTMC
X(t)

Two major components:

 Ti = sojourn time in state i (random variable)
 pij = probability of moving to state j when leaving state i
Continuous Time Markov Chain
Sojourn time in a state
 Let Ti be the random variable corresponding to the
time spent in state i
 The memoryless property of the homogenuous CTMC
implies
P Ti  t  x ¨Ti  t   P Ti  x , "t , "x

 The exponential distribution is the only continuous

probability distribution having this property.

In an CTMC, the sojourn time in any state is exponentially

distributed.
 Continuous Time Markov Chain
 Let T be a continuous random variable with an exponential
distribution of parameter 
 Distribution Function (figure) :
1  e t , t  0
FT(t) = P{T ≤ t} FT  t   
 0, t0

 Probability density function :  et , t  0

fT  t   
fT(t) = dFT(t)/dt 0, t0
 Mean : E[T] = 1/ l
 Standard deviation: s[T] = 1/ l
 Coeficient of variation: Cv(T) = s[T]/ E[T] = 1
 Parameter  often corresponds to some event rate (failure
rate, repair rate, production rate, ...)
Continuous Time Markov Chain
fT(t)
The Exponential Distribution

Probability density

t
Mean=
Time between arrivals
E[T]=1/
Continuous Time Markov Chain
 When a CTMC leaves state i, it jumps to state j with
probability pij. This probability is:
 independent of time as the CTMC is homogeneous
 independent of sojourn time Ti as the process is
markovian (memoryless)
Continuous Time Markov Chain
1st characterization of a CTMC
 An CTMC is fully characterized by the following
parameters:
 {mi}iE with mi as the parameter of the exponential
distribution of sojourn time Ti
 {pij}i≠j , with pij as the transition probability from i to j
when leaving state i
Continuous Time Markov Chain
 Each CTMC is associated an underlying DTMC by
neglecting sojourn times.

 A state i of a CTMC is said transient (resp. recurrent,

absorbing) if it is transient (resp. recurrent, absorbing)
in the underlying DTCM

 A CTMC is irreducible if its underlying DTMC is

irreducible.

 Remark: the concept of periodicity is not relevant.

Continuous Time Markov Chain
2nd characterization of a CTMC
 Each state activates several potential events leading
to different transitions.
 A CTMC travels from state i to state j in Tij time, an
exponentially distributed random variable with
parameter ij.
 i is called transition rate from i to j.
 Birth-death Process in Queuing
 A birth-death (BD process) process refers to a Markov
process with
 a discrete state space
 the states of which can be enumerated with index i=0,1,2,. . .
such that
 state transitions can occur only between neighboring states, i
→ i + 1 or i → i − 1
 Birth-death Process in Queuing
 The foundation of many of the most commonly used
queuing models
 Birth – equivalent to the arrival of a customer or job
 Death – equivalent to the departure of a served customer or
job
 Pure birth process is originated from Poisson process
in which the rate of birth depends on number of
persons.
 In a birth-death system once serviced a customer
moves to the next state.
 This is like a non-deterministic finite-state machine.
 Birth-death Process in Queuing
Pure Birth Process; e.g., Hurricanes

0
a0
1
a1
2
a2
3
a3
4 …

Pure Death Process; e.g., Delivery of a truckload of parcels

0
d1
1
d2
2
d3
3
d4
4 …

Birth-Death Process; e.g., Repair shop for taxi company

d1 d2 d3 d4
0 1 2 3 4 …
a0 a1 a2 a3
 Queuing Models
A Commonly Seen Queuing Model

The Service Facility

C S = Server
The Queue C S
Customers (C)
CCC…C •
•
•

Customer =C C S
 Queuing Models
 Service times as well as inter-arrival times are
assumed independent and identically distributed
 If not otherwise specified
 Commonly used notation principle: A/B/C
 A = The interarrival time distribution
 B = The service time distribution
 C = The number of parallel servers
 Commonly used distributions
 M = Markovian (exponential) - Memoryless
 D = Deterministic distribution
 G = General distribution
 Example: M/M/c
 Queuing system with exponentially distributed service and
inter-arrival times and c servers
 Queuing Models
Notation of Kendall
 The following is a standard notation system of
queueing systems
 T/X/C/K/P/Z with
 T: probability distribution of inter-arrival times
 X: probability distribution of service times
 C: Number of servers
 K: Queue capacity
 P: Size of the population
 Z: service discipline
 Queuing Models
Customer arrival process : T/X/C/K/P/Z

 T can take the following values:

 M : markovian (i.e. exponential)
 G : general distribution
 D : deterministic
 Ek : Erlang distribution
 If the arrivals are grouped in lots, we use the notation T[X]
where X is the random variable indicating the number of
customers at each arrival epoch
 P{X=k} = P{k customers arrive at the same time}
 Some arriving customers can leave if the queue is too long
 Queuing Models
Service times : T/X/C/K/P/Z
• X Can take the following values:
 M : markovian (i.e. exponential)
 G : general distribution
 D : deterministic
 Ek : Erlang distribution

k exponential servers with parameter m

Erlang distribution Ek with parameter 

 Queuing Models
Number of servers : T/X/C/K/P/Z
 In simple queueing systems, servers are identical
 Queuing Models
Queue capacity : T/X/C/K/P/Z

Loss of customers if
the queue Is full

Capacity K
 Queuing Models
Size of the population : T/X/C/K/P/Z
 The size of the population can be either finite or
infinite

 For a finite population, the customer arrival rate is a

function of the number of customers in the system:
(n).
 Queuing Models
Service discipline : T/X/C/K/P/Z
 Z can take the following values:
 FCFS or FIFO : First Come First Served
 LCFS or LIFO : Last Come First Served
 RANDOM : service in random order
 HL (Hold On Line) : when an important customer arrives, it
takes the head of the queue
 PR ( Preemption) : when an important customer arrives, it is
served immediately and the customer under service returns to
the queue
 PS (Processor Sharing) : All customers are served
simultaneously with service rate inversely proportional to the
number of customers
 GD (General Discipline)
 Queuing Models
Simplified notation
 We will use the simplified notation T/X/C when we
consider a queue where:
 The capacity is infinite
 The size of the population is infinite
 The service discipline is FIFO

 Hence T/X/C = T/X/C///FIFO

 Poisson Process
 The standard assumption in many queuing models is
that the arrival process is Poisson
 The Poisson distribution is often used for random
arrivals
In Poisson Distribution:
 Average arrival rate is known
 Average arrival rate is constant for some number of
time periods
 Number of arrivals in each time period is independent
 As the time interval approaches 0, the average
number of arrivals approaches 0
 Poisson Process

 Poisson Process
Properties of the Poisson Process
 Poisson processes can be aggregated or
disaggregated and the resulting processes are also
Poisson processes
a) Aggregation of N Poisson processes with intensities
{1, 2, …, n} renders a new Poisson process with
intensity = 1+ 2+…+ n.
b) Disaggregating a Poisson process X(t)Po(t) into N
sub-processes {X1(t), X2(t), , …, X3(t)} (for example N
customer types) where Xi(t) Po(it) can be done if
 For every arrival the probability of belonging to sub-
process i = pi
 p1+ p2+…+ pN = 1, and i = pi 
 Terminology and Notation
 The state of the system = the number of customers in the
system
 Queue length = (The state of the system) – (number of
customers being served)
N(t) = Number of customers/jobs in the system at time t
Pn(t) = The probability that at time t, there are n customers/jobs in the
system.
n = Average arrival intensity (= # arrivals per time unit) at n
customers/jobs in the system
n = Average service intensity for the system when there are n
customers/jobs in it. (Note, the total service intensity for all occupied
servers)
 = The utilization factor for the service facility. (= The expected
fraction of the time that the service facility is being used)
 Terminology and Notation
 Consider an M/M/1 queue with arrival rate =  and service
intensity = 
  = Expected capacity demand per time unit
  = Expected capacity per time unit
Capacity Demand λ
ρ  
Available Capacity μ

 Similarly if there are c servers in parallel, i.e., an M/M/c

system but the expected capacity per time unit is then c*

Capacity Demand 
 

Available Capacity c * 
 Terminology and Notation
 Steady State condition
 Enough time has passed for the system state to be independent
of the initial state as well as the elapsed time
 The probability distribution of the state of the system remains the
same over time (is stationary).
 Transient condition
 Prevalent when a queuing system has recently begun operations
 The state of the system is greatly affected by the initial state and
by the time elapsed since operations started
 The probability distribution of the state of the system changes
with time

With few exceptions Queuing Theory has focused on

analyzing steady state behavior
Notation For Steady State Analysis
Pn = The probability that there are exactly n
customers/jobs in the system (in steady state, i.e.,
when t)
L = Expected number of customers in the system (in
steady state)
Lq = Expected number of customers in the queue (in
steady state)
W = Expected time a job spends in the system
Wq= Expected time a job spends in the queue
 Little’s Formula Revisited
 Assume that n =  and n =  for all n

L  W Lq  Wq

 Assume that n is dependent on n


Let    Pn  n
n 0

L  W Lq   Wq
 Queuing Models: M/M/1
 M/M/1 queue is a useful approximate model when
service times have standard deviation approximately
equal to their means.
 Single Server Queuing System (M/M/1)
 Only one server
 Poisson arrivals
 Arrival population is unlimited
 Exponential service times
 All arrivals wait to be served
 λ is constant
 μ > λ (average service rate > average arrival rate)
 Queuing Models: M/M/1
Assumptions - the Basic Queuing Process
 Infinite Calling Populations
 Independence between arrivals
 The arrival process is Poisson with an expected arrival
rate 
 Independent of the number of customers currently in the
system
 The queue configuration is a single queue with possibly
infinite length
 No reneging or balking
 The queue discipline is FIFO
 The service mechanism consists of a single server with
exponentially distributed service times
  = expected service rate when the server is busy
 Queuing Models: M/M/1

 Queuing Models: M/M/1
 n=  and n = for all values of n=0, 1, 2, …
   

0 1 2  n-1 n n+1

   

 Average server utilization:  = (/) < 1

 Average number of customers waiting: Lq= 2/(1- )
= 2/ ( - )
 Average number of customers in system: L= Lq +  or
L=/(1- )
 Average waiting time: Wq=Lq/=  /( (- ))
 Average waiting time in system: W = Wq + 1/  or
W=L/=1/(- )
 Queuing Models: M/M/1
 n=  and n = for all values of n=0, 1, 2, …
   

0 1 2  n-1 n n+1

   

 Probability of 0 customers in system: P0 = 1- = 1 - (/)

 Probability of exactly n customers in system:

Pn = n(1- )
= (λ/μ )n P0
 Queuing Models: M/M/1
 Example: M/M/1 queue with service rate m=10
customers per hour.
 Consider how L and w increase as arrival rate, l, increases
from 5 to 8.64 by increments of 20%:

 5.0 6.0 7.2 8.64 10.0

 0.500 0.600 0.720 0.864 1.000
L 1.00 1.50 2.57 6.35 ∞
w 0.20 0.25 0.36 0.73 ∞
 If /  1, waiting lines tend to continually grow in length.
 Increase in average system time (w) and average number in
system (L) is highly non-linear as a function of .
 Queuing Models: M/M/1
 Example: If arrivals are occurring at rate  = 10 per
hour. Management has a choice of two servers, one
who works at rate 1=11 customers per hour and
second at rate 2=12 customers per hour.
Respective utilizations are:
 1=/1=10/11=0.909; 2=/2=10/12=0.833
Average number in the system is:
 L1= 1/(1- 1)=10; L2= 2/(1- 2)=5
Thus, a decrease in service rate from 12 to 11
customers per hour, a mere 8.3% decrease would result
in an increase in average number in system from 5 to 10,
which is 100% increase!!!
 Queuing Models: M/M/C
 Multiple Server System (M/M/C)
 Poisson arrivals
 Exponential service times
 C servers
 Total service rate must exceed arrival rate ( cμ > λ)
 Many of the operating characteristic formulas are more
complicated
 Queuing Models: M/M/C
 Generalization of the M/M/1 model
– Allows for c identical servers working independently from each other
    

0 1 2  c-2 c-1 c c+1

 2 (c-2) (c-1) c c

1
 c1 ( / ) n ( / )c
1 
P0      
 n! c! 1  ( /(c) 
 n 0
Steady State
 ( /  ) n Condition:
 P0 for n  1,2,, c
 n! =(/c)<1
Pn  
 ( /  ) P
n
for n  c  1, c  2,
 c!c n c 0
 Queuing Models: M/M/C
 A Condition for existence of a steady state solution is that
 = /(c) <1

 ( /  ) c 
Lq   ( n  c) Pn  ...  P
2 0
n c c!(1  )

Little’s Formula  Wq=Lq/

W=Wq+(1/)

Little’s Formula  L=W= (Wq+1/ ) = Lq+ / 

 Queuing Models: M/G/1
Single Server System With General Service Time
(M/G/1)
 Poisson arrivals
 General service time distribution with known mean (μ)
and standard deviation (σ)
 μ>λ
 Single-server queues with unlimited capacity.
 Queuing Models: M/G/1
 Suppose service times have mean 1/m and variance s2
and r = l/m < 1, the steady-state parameters of M/G/1
queue:
   /  , P0  (1   ) is the probabilit y server is idle (no customers)
2 (1  2   2 )  2 (1   2  2 )
L 
2(1   ) 2(1   )
 2 (1   2  2 )
LQ 
2(1   )
 (1 /  2   2 )
1
w 
 2(1   )
 (1 /  2   2 )
wQ 
2(1   )
 Queuing Models: M/G/1
 No simple expression for the steady-state probabilities
P0, P1, …
 L – LQ =  is the time-average number of customers
being served.
 Average length of queue, LQ, can be rewritten as:
2 2 2
LQ  
2(1   ) 2(1   )

 If  and  are held constant, LQ depends on the variability,

2, of the service times.
 Queuing Models: M/G/1
 Example: Widget making machines malfunction at random
and then require mechanic’s attention. Malfunctions occur
according to a poisson process at a rate l = 1.5 per hour.
Observation over several months has found that repair times
by the single mechanic take an average time of 30 minutes
with a standard deviation of 20 minutes.
Therefore, the mean service time is 1/m = ½ =0.5 hour, the
service rate m = 2 per hour and the variance s2 = 202
minutes = 1/9 hour.
The customers are the widget makers and appropriate
model is M/G/1 because only the mean and variance of the
service times are known, not the distribution
Therefore, =/=1.5/2 = 0.75.
Steady state time average number of broken machines is:
(1.5) 2 [0.52  1 / 9]
L  0.75   2.375 customers
2(1  0.75)
 Queuing Models: M/G/1
 Example: Two workers competing for a job, Able claims to be
faster than Baker on average, but Baker claims to be more
consistent,
 Poisson arrivals at rate l = 2 per hour (1/30 per minute).
 Able: 1/m = 24 minutes and s2 = 202 = 400 minutes2
(1 / 30) 2 [24 2  400 ]
LQ   2.711 customers
2(1  4 / 5)

The proportion of arrivals who find Able idle and thus experience no
delay is P0 = 1- = 1/5 = 20%.
 Baker: 1/ = 25 minutes and 2 = 22 = 4 minutes2:
(1 / 30) 2 [25 2  4]
LQ   2.097 customers
2(1  5 / 6)
The proportion of arrivals who find Baker idle and thus experience
no delay is P0 = 1- = 1/6 = 16.7%.
Although working faster on average, Able’s greater service variability
results in an average queue length about 30% greater than Baker’s.
 Queuing Models: M/D/1
Single Server System With Constant Service Time
(M/D/1)
 Poisson arrivals
 Service times deterministic and constant (not random)
 Has shorter queues than M/M/1 system
 Lq and Wq are one-half as large
 Queuing Models: M/D/1
 Average length of queue
2
Lq 
21   
 Average time waiting in queue
1   
Wq   
2  1   
 Average time spent in system
1 2  
W  
2  1   
 Queuing Models: M/D/1

Queuing Modeling and System Design
 Design of queuing systems usually involve some kind
of capacity decision
 The number of service stations
 The number of servers per station
 The service time for individual servers
The corresponding decision variables are , c and 

 Examples:
 The number of doctors in a hospital,
 The number of exits and cashiers in a supermarket,
 The choice of machine type at a new investment decision,
 The localization of toilets in a new building, etc…
Queuing Modeling and System Design
 Two fundamental questions when designing (queuing)
systems
 Which service level should we aim for?
 How much capacity should we acquire?
 The cost of increased capacity must be balanced
against the cost reduction due to shorter waiting time
 Specify a waiting cost or a shortage cost accruing when
customers have to wait for service or…
 … Specify an acceptable service level and minimize the
capacity under this condition
 The shortage or waiting cost rate is situation dependent
and often difficult to quantify
 Should reflect the monetary impact a delay has on the
organization where the queuing system resides