Wireless Networking Fundamentals

Rajasekar Mohan
Department of Electronics and Communication Engineering
[email protected]

Prof. Rajasekar Mohan 1

Wireless Networking Fundamentals

Unit 4: Traffic Engineering I

Prof. Rajasekar Mohan

Department of Electronics and Communication Engineering
Prof. Rajasekar Mohan 2
Queuing Concept

Prof. Rajasekar Mohan 3

Queuing Theory

Queuing theory

Queuing theory is the mathematical study of waiting lines, or

queues.

A queuing model is constructed so that queue lengths and

waiting time can be predicted.

Creates an efficient and cost-effective workflow, allowing the

user to improve traffic flow

Prof. Rajasekar Mohan 4

Applications of Queuing theory

Telecommunications Transportation
Reliability Engineering Logistics
Computing Finance
Industrial Engineering Emergency Services
Project Management

Prof. Rajasekar Mohan 5

Importance of Queuing Theory
Queuing models calculate the best number of servers to
minimize costs.
Queue lengths and waiting times can be predicted.
Improve customer Service, continuously.
To mitigate congestion in the network
Queuing theory provides all the tools needed for the
mathematical analysis of network flow.
To build systems for latency critical applications.
Estimate the performance measures.

6
Performance Measures

• Delay

• Mean queue size,

• Blocking probability and

• Throughput

7
Performance Measures

8
Performance Measures

9
Performance Measures

10
Performance Measures

11
Performance Measures

12
Blocking Probability
Blocking Probability: Probability of rejecting a customer

13
Queue size or length

14
Blocking Probability
Blocking Probability: Probability of rejecting a customer

15
Blocking Probability
Blocking Probability: Probability of rejecting a customer

16
Utilisation

17
Throughput

18
Stochastic process

For a given index set T, a stochastic process {Xt, t ∈ T} is an

indexed collection of random variables.

If the index set T is countable, the stochastic process is called

a discrete-time process, or a time series . Otherwise, the
stochastic process is called a continuous-time process.

A stochastic process {Xt, t ∈ T} is called discrete space stochastic

process if the random variables Xt are discrete, and it is called
continuous space stochastic process if it is continuous.

If a stochastic process is stationary, then the probability of equal

length random sequences will remainProf.the same
Rajasekar Mohan 19
Types of stochastic processes

Discrete Time Discrete Space

 Discrete Time Continuous Space

 Continuous Time Discrete Space

 Continuous Time Continuous Space.

Prof. Rajasekar Mohan 20

Gaussian process

A process that has the property that the joint probability

function (density) associated with any set of times is
multivariate Gaussian.

Gaussian process provides an accurate model for

superposition of many independent random processes

Example: Heavily multiplexed traffic which arrive at

switches or routers deep in a major computer network
are described as Gaussian process

Gaussian process is simple and amenable to analysis.

21
Point process

A point process is a sequence of events called as arrivals

occurring at random in points of time ti, i = 1, 2, . . . , ti+1 >
ti, or i =. . . , − 2,−1, 0, 1, 2, . . ., ti+1 > ti.

A point process can also be defined as a counting process

denoted by {N(t), t ≥ 0} N(t) represents the number of
arrivals up to time t.

Prof. Rajasekar Mohan 22

Point process

A point process obeys two properties:

• Oderliness and
• memorylessness.

Two processes qualify the above definition:

• Bernoulli process (discrete-time)

• Poisson process (continuous-time)

Prof. Rajasekar Mohan 23

Bernoulli process
The Bernoulli process is a discrete-time stochastic process made
up of a sequence of IID Bernoulli distributed random variables
{Xi, i = 0, 1, 2, 3, . . .} where for all i, P(Xi = 1) = p and P(Xi = 0) =
1−p.

The Bernoulli process is both orderly and memory less

The counting process {N(n), n ≥ 0} for the Bernoulli process is

another random process of Binomial random variables N(n)
where N(n) represents the total number of arrivals occurring
within the first n time-slots.

Prof. Rajasekar Mohan 24

Bernoulli process

Inter-arrival times (∆) in a Bernoulli process are IID.

Inter-arrival times represents the number of Bernoulli trials

until a success

Hence, ∆ follows geometric distribution

Bernoulli process is preserved by superposition, merging

and splitting

Prof. Rajasekar Mohan 25

Poisson process

A counting process {N(t), t ≥ 0} is defined as a Poisson

process with rate λ > 0 if it satisfies the following:

• N(0) = 0
• The number of occurrences in two non-overlapping
intervals are independent.
• The number of occurrences in an interval of length t
has a Poisson distribution with mean λt

Inter-arrival times of the occurrences in Poisson process

are exponentially distributed with parameter λ

Prof. Rajasekar Mohan 26

Properties of Poisson point process

Expanding time series

Prof. Rajasekar Mohan 27

Properties of Poisson point process

Prof. Rajasekar Mohan 28

Properties of Poisson point process

Prof. Rajasekar Mohan 29

Poisson process

• Poisson process with parameter λ is denoted by

• The mean inter-arrival time is expressed as

• The number of occurrences in interval T is given by λT

Prof. Rajasekar Mohan 30

Poisson process

Supports superposition, merging and splitting

If a Poisson stream is split into k sub-streams with probability
pi, then each sub-stream is also Poisson with mean piλ

Prof. Rajasekar Mohan 31

Poisson process

Merging of k Poisson streams with mean λi results in a

Poisson with mean rate (λ1+ λ2+...+ λk)

Prof. Rajasekar Mohan 32

Markov chains

 Discrete space stochastic processes

 Popular for analysis, traffic characterization and modelling of
queuing and telecommunications networks and systems.
 The evolution of a point/counting process can be defined by its
state probabilities and state transition probabilities.
 They can be classified into two groups:
• Discrete-time Markov-chains
• Continuous time Markov-chains

Prof. Rajasekar Mohan 33

Graph of a Markov Chain

A diagram representing a two-state Markov

process, with the states labelled E and A.
Each number represents the probability of the
Markov process changing from one state to
another state, with the direction indicated by
the arrow. For example, if the Markov process
is in state A, then the probability it changes to
state E is 0.4, while the probability it remains
in state A is 0.6.

Prof. Rajasekar Mohan 34

Example of a Markov Chain
Dietary habits of a creature who eats only grapes, cheese, or lettuce,
and whose dietary habits conform to the following rules:

• It eats exactly once a day.

• If it ate cheese today, tomorrow it will eat lettuce or grapes with equal
probability.
• If it ate grapes today, tomorrow it will eat grapes with probability 1/10,
cheese with probability 4/10, and lettuce with probability 5/10.
• If it ate lettuce today, tomorrow it will eat grapes with probability 4/10
or cheese with probability 6/10. It will not eat lettuce again tomorrow.

This creature's eating habits can be modeled with a Markov chain

since its choice tomorrow depends solely on what it ate today, not
what it ate yesterday or any other time in the past. One statistical
property that could be calculated is the expected percentage, over a
long period, of the days on which the creature will eat grapes.
Prof. Rajasekar Mohan 35
Markov Chain

A Markov chain is a mathematical system that experiences

transitions from one state to another according to
certain probabilistic rules. The defining characteristic of a Markov
chain is that no matter how the process arrived at its present state,
the possible future states are fixed. In other words, the probability of
transitioning to any particular state is dependent solely on the
current state and time elapsed. The state space, or set of all possible
states, can be anything: letters, numbers, weather conditions,
baseball scores, or stock performances.

Prof. Rajasekar Mohan 36

Markov Chain

Markov chains may be modeled by finite state machines, and random

walks provide a prolific example of their usefulness in mathematics.
They arise broadly in statistical and information-theoretical contexts
and are widely employed in economics, game theory, queueing
(communication) theory, genetics, and finance.

While it is possible to discuss Markov chains with any size of state

space, the initial theory and most applications are focused on cases
with a finite (or countably infinite) number of states.

Many uses of Markov chains require proficiency with

common matrix methods.

Prof. Rajasekar Mohan 37

Example of a Markov Chain

Prof. Rajasekar Mohan 38

Example of a Markov Chain
A (time-homogeneous) Markov chain built on states A and B is depicted in
the diagram below. What is the probability that a process beginning on A will
be on B after 2 moves?
In probability theory,
the most immediate
example is that of
a time-homogeneous
Markov chain, in which
the probability of any
state transition is
independent of time.
Such a process may be
visualized with a
labeled directed graph,
for which the sum of
the labels of any
vertex's outgoing
edges is 1.
Prof. Rajasekar Mohan 39
Example of a Markov Chain
A (time-homogeneous) Markov chain built on states A and B is depicted in
the diagram below. What is the probability that a process beginning on A will
be on B after 2 moves?
In probability theory,
the most immediate
example is that of
a time-homogeneous
Markov chain, in which
the probability of any
In order to move from A to B, the process must either stay on A the first state transition is
move, then move to B the second move; or move to B the first move, independent of time.
then stay on B the second move. Such a process may be
According to the diagram, the probability of that is  visualized with a
labeled directed graph,
0.3 x 0.7 + 0.7 x 0.2 = 0.21+0.14 = .35​ for which the sum of
the labels of any
vertex's outgoing
edges is 1.
Prof. Rajasekar Mohan 40
Example of a Markov Chain
A (time-homogeneous) Markov chain built on states A and B is depicted in
the diagram below. What is the probability that a process beginning on A will
be on B after 2 moves?
In probability theory,
the most immediate
example is that of
a time-homogeneous
Markov chain, in which
the probability of any
Alternatively, the probability that the process will be on A after 2 moves state transition is
is 0.3 x 0.3 + 0.7 x 0.8 = 0.65 independent of time.
Such a process may be
Since there are only two states in the chain, the process must be on B if visualized with a
it is not on A, and therefore the probability that the process will be on B labeled directed graph,
after 2 moves is 1 - 0.65 = 1−0.65=0.35​. for which the sum of
the labels of any
vertex's outgoing
edges is 1.
Prof. Rajasekar Mohan 41
Markov Property

In other words, knowledge of the previous state is all that is

necessary to determine the probability distribution of the current
state.

This definition is broader, as it allows for non-stationary transition

probabilities and therefore time-inhomogeneous Markov chains; 

that is, as time goes on (steps increase), the probability of

moving from one state to another may change.
Prof. Rajasekar Mohan 42
Discrete time Markov chains

 A discrete-time Markov-chain is a discrete-time stochastic

process {Xn, n = 0, 1, 2, . . .} with the Markov property

The state of the process can be a scalar or a vector

Prof. Rajasekar Mohan 43

Transition probability

 Transition probability matrix P is a matrix of one-step transition

probabilities

 Sum of elements of each row equals

 The n-step transition probabilities is given by

 Also known as Chapman-Kolmogorov equations

Prof. Rajasekar Mohan 44
Transition probability

Prof. Rajasekar Mohan 45

Transition probability

Prof. Rajasekar Mohan 46

Transition probability

Prof. Rajasekar Mohan 47

Transition probability
Once people arrive in Thailand, they want to enjoy the sun
and beaches on 2 popular islands in the south: Samui
Island & Phangan Island.

From survey data, when on the mainland, 70% of tourists

plan to go to Samui Island, 20% to Phangan Island, and
only 10% remain on shore the next day.

When on Samui Island, 40% continue to stay on Samui,

50% plan to go to Phangan Island, and only 10% return to
mainland the next day.

Finally, when on Phangan Island, 30% prolong their stay

here, 30% divert to Samui Island, and 40% go back to
mainland the next day.

Starting from the mainland, what is the probability (in

percentage) that the travelers will be on the mainland at
the end Prof.
of aRajasekar
3-dayMohan
trip? 48
Transition probability

For the time-independent

Markov chain described by the
picture below, what is its 2-step
transition matrix?

Prof. Rajasekar Mohan 49

 Transition probability
A Markov chain has first state A
and second state B, and its
transition probabilities for all
time are given by the following
graph:

What is its transition matrix? Note: The transition matrix is oriented such that

the kth row represents the set of probabilities of
transitioning from state k to another state.

Prof. Rajasekar Mohan 50

Markov Properties

Prof. Rajasekar Mohan 51

 Markov Properties

Determine facts
about their
periodicity and
reducibility.
Prof. Rajasekar Mohan 52
 Markov Properties
In Markov Chain 1, there is no way to
get from state A (or C) to state B. Any
move from state A (or C) to itself must
take an even number of steps, so that
state is periodic. Thus, Chain 1 is not The First Markov chain does not
irreducible and not aperiodic. allow transition to state B from
In Markov Chain 2, there is a way to state A or C though its
get from any given state to any other. probabilities are positive. That
Any move from state A (or B, or C) to means it's not irreducible. It is
itself must take a number of steps also taking more than 1 step to
divisible by 3, so that state is periodic. transit from state A to State C or
Thus, Chain 1 is irreducible and not state C to State A, so it cannot
aperiodic. be aperiodic. So it periodic
Markov Chain.
In Markov chain two, it's
Determine facts possible to transition from one
about their state to other with three move
periodicity and and positive probability, so it's
reducibility. not aperiodic and irreducible.
Prof. Rajasekar Mohan 53
Marginal Distribution
Suppose there is data from classroom of 200 students on the amount of time studied (X) and the
percent correct (Y).[4] Assuming that X and Y are discrete random variables, the joint distribution
of X and Y can be described by listing all the possible values of p(xi,yj), as shown in Table.

Prof. Rajasekar Mohan 54

Marginal Distribution
The marginal distribution can
be used to determine how
many students that scored 20
or below

The conditional distribution can be used to determine

the probability that a student scored 20 or below while
also studying for 60 minutes or more

meaning there is about a 11% probability of scoring 20 after having studied for at least 60 minutes.

Prof. Rajasekar Mohan 55

Marginal probability

Marginal probabilities are given by

Prof. Rajasekar Mohan 56

 Steady state probability

Steady state probabilities

Markov-chain is irreducible if all the states in its state

space are accessible from each other.
Prof. Rajasekar Mohan 57
Steady state probability

Prof. Rajasekar Mohan 58

Continuous time Markov Chain

At any point in time t, {Xt} describes the state of the process
which takes a discrete value, where Xt takes on one of the
nonnegative integer values (0,1,2,....L)

The CTMC describes the evolution of a continuous time

stochastic process which follows Markov property

where

Prof. Rajasekar Mohan 59

Continuous time Markov Chain

Given one of the L states that Xt takes (say state i),

transition from state i to some state j ϵ {0,1, 2, ..., L}\i
occurs at the rate qij

• In other words, the state i changes to state j after

some random time based on the exponential
distribution with mean rate qij
• Imagine there are L-1 clocks corresponding to each
state in {0,1, 2, ...., L}\i, started when state i occurs.
Each clock randomly expires. The first clock to
expire determines the next state j.
• At every state transition, all clocks are reset
Prof. Rajasekar Mohan 60
Continuous time Markov Chain
The amount of time spent in state i (denoted by Ti) is exponentially
distributed with parameter δi before making a transition into a
different state

where

Transition probability

Prof. Rajasekar Mohan 61

CTMC: Steady state probabilities
Steady state probabilities are denoted by Π
The relation between Q and Π is given by
ΠQ = 0
where, Q is the infinitesimal generator matrix whose
elements are given by

Rate of transition from state i to state j

= rate to leave state i × probability of state transition from i to j

Prof. Rajasekar Mohan 62

CTMC: Steady state probabilities
(contd.)

Expanding the above expression for the j-th row of ΠQ = 0,

Prof. Rajasekar Mohan 63

CTMC: Steady state probabilities
(contd.)

Global balance equations mean that the expected

number of transitions in and out of state j must be equal
In other words, the probability flux going in and out of
state j must be equal
Using ΠQ = 0 and the normalization condition

steady state distribution Π can be found out.

Prof. Rajasekar Mohan 64
Queuing and Teletraffic Concepts

Queuing systems may be characterized by

complex input process

 service time distribution

number of servers (or channels)

buffer size (or waiting room) and

queue disciplines.

Prof. Rajasekar Mohan 65

 Notations
The following notation is used for representing queues:

A/B/c/K
where
• A denotes the distribution of the inter-arrival time,
• B that of the service time,
• c denotes the number of servers, and
• K denotes the capacity of the queue.

If K is omitted, we assume that K = 1.

M stands for Markov and is commonly used for the exponential distribution.

Hence an M/M/1 queue is one in which there is one server (and one channel) and both the inter- arrival time
and service time are exponentially distributed.

An M/G/1 queue is one with one server in which the inter-arrival time is exponentially distributed and the
service time is generally distributed, i.e., the service time has any given distribution. A G/G/1 queue is one with
one server in which both service and the inter-arrival Prof. time have
Rajasekar any given distribution.
Mohan 66
Kendall notation

Kendall notation: {arrival process} / {service

distribution} / {number of servers} / {buffer size} -
{queue discipline}

Notation for first two positions:

D (Deterministic), M (Markovian), G (General),
GI (General and independent), and Geom
(Geometric)

Prof. Rajasekar Mohan 67

Kendall Notation of Queuing Systems

Prof. Rajasekar Mohan 68

Kendall Notation of Queuing Systems

Prof. Rajasekar Mohan 69

Queuing disciplines

First In First Out (FIFO), Last In First Out (LIFO),

Processor Sharing (PS) where all customers in the
queue obtain service, and random order
(random)
Service facility can have series or parallel servers
Service facility can have finite or infinite buffer
• Buffer holds workloads awaiting service
Service capacity / buffer size + no. of servers

Prof. Rajasekar Mohan 70

Queuing Model

Workload
Waiting workloads being processed

Arrival Departure

Service facility Server

Prof. Rajasekar Mohan 71

System Utilization

It is a measure of how busy a system is compared to the

total time it is available

It can be expressed as the ratio of the ON time to ON

time + OFF time

It is denoted by ρ

Prof. Rajasekar Mohan 72

Traffic intensity

It is a measure of the total arrival presented to a system

as a whole

Traffic intensity
= mean arrival rate × mean service time =
λ/μ (Erlangs)

Prof. Rajasekar Mohan 73

 Traffic intensity
Consider a wireless system that provides channels each of which can serve one phone call. There
are 100 users making phone calls. Each user makes on average one phone call per hour and the
average duration of a phone call is three minutes. What is the total traffic in erlangs that the
100 users generate.

Prof. Rajasekar Mohan 74

 Traffic intensity
The quantity of traffic measured in erlangs is also called traffic intensity. Then another related
concept is traffic volume. Traffic volume is measured in units of erlang-hour (or erlang-minute, or
call-hour or call-minute, etc.) and it is a measure of the traffic processed by a facility during a given
period of time. Traffic volume is the product of the traffic intensity and the given time period, namely,

Prof. Rajasekar Mohan 75

Traffic intensity

Prof. Rajasekar Mohan 76

Traffic intensity

Prof. Rajasekar Mohan 77

Little’s Theorem

The average number of customers (Q) in a queuing

system is equal to the average arrival rate (λ) of
customers to that system times the average time (D)
a customer spent in that system

Little’s theorem: Q = λ×D

Prof. Rajasekar Mohan 78

 Work Conservation

• Another important concept in queuing theory is the concept of work

conservation.

• A queuing system is said to be work conservative if a server is

never idle whenever there is still work to be done.

• For example, M/M/1 and M/M/1/7 are work conservative.

• However, a stable M/M/3 is not work conservative because a server

can be idle while there are customers served by other servers.

Prof. Rajasekar Mohan 79

Flow Conservation Principle

In a stable system (a system in steady-state), what

goes in must go out in a finite amount of time.

Prof. Rajasekar Mohan 80

Customer’s Behaviour

Prof. Rajasekar Mohan 81

Birth-and-Death Process

Prof. Rajasekar Mohan 82

M/M/1 model – Single server model

Arrival process is Poisson with parameter λ

Service times are IID and assumed to be exponentially
distributed with parameter μ
At t = 0, the state of the system is 0
After a random period based on exponential
distribution with mean 1/λ, the state changes to 1
The time the process stays in state n (where n ≥1) is
also exponentially distributed

Prof. Rajasekar Mohan 83

M/M/1 model – Single server model

Arrival process is Poisson with parameter λ

Service times are IID and assumed to be exponentially
distributed with parameter μ
At t = 0, the state of the system is 0
After a random period based on exponential
distribution with mean 1/λ, the state changes to 1
The time the process stays in state n (where n ≥1) is
also exponentially distributed

Prof. Rajasekar Mohan 84

 M/M/1 model
(contd.)

The state n may last till the time of next arrival

(exponentially distributed with λ)
Alternatively, the state n may last till the time
of next departure (exponentially distributed
with μ)
The minimum of the above two is
exponentially distributed with λ+μ
Hence, this is the time the process stays in
state n (where n ≥1)

Prof. Rajasekar Mohan 85

M/M/1 model
(contd.)
The probability of the state transition from state n
to n+1 is given by λ/(λ+μ)
The probability of the state transition from state n
to n–1 is given by μ /(λ+μ)
The M/M/1 queue-size process increases by only
one, decreases by only one and stays an
exponential amount of time at each state.
Two variations of the M/M/1 model are
considered:
• infinite buffer and
• finite buffer
Prof. Rajasekar Mohan 86
M/M/1 model

Prof. Rajasekar Mohan 87

M/M/1 model

Prof. Rajasekar Mohan 88

M/M/1 model - Example

Prof. Rajasekar Mohan 89

M/M/1 model - Example

Prof. Rajasekar Mohan 90

What-If an Extra employee is added

Prof. Rajasekar Mohan 91

What-If an Extra Counter & Employee are added

Prof. Rajasekar Mohan 92

Example

Consider the following single-server queue: the inter-arrival time is exponentially distributed with
a mean of 10 minutes and the service time is also exponentially distributed with a mean of 8
minutes, Find the (i) mean wait in the queue, (ii) mean number in the queue, (iii) the mean
wait in the system, (iv) mean number in the system and (v) proportion of time the server is
idle.

Prof. Rajasekar Mohan 93

Consider the following single-server queue: the inter-arrival time is exponentially distributed with
a mean of 10 minutes and the service time is also exponentially distributed with a mean of 8
minutes, Find the (i) mean wait in the queue, (ii) mean number in the queue, (iii) the mean
wait in the system, (iv) mean number in the system and (v) proportion of time the server is
idle.

Prof. Rajasekar Mohan 94

Example

Consider the following single-server queue: the inter-arrival time is exponentially distributed
with a mean of 10 minutes and the service time has the uniform distribution with a
maximum of 9 minutes and a minimum of 7 minutes, find the (i) mean wait in the queue, (ii)
mean number in the queue, (iii) the mean wait in the system, (iv) mean number in the
system and (v) proportion of time the server is idle.

Prof. Rajasekar Mohan 95

 Consider the following
single-server queue: the
inter-arrival time is
exponentially distributed
with a mean of 10 minutes
and the service time has
the uniform distribution
with a maximum of 9
minutes and a minimum of
7 minutes, Find the (i)
mean wait in the queue, (ii)
mean number in the
queue, (iii) the mean wait
in the system, (iv) mean
number in the system and
(v) proportion of time the
server is idle.

Prof. Rajasekar Mohan 96

Example

Prof. Rajasekar Mohan 97

Example

Prof. Rajasekar Mohan 98

Steady-State Probabilities

Prof. Rajasekar Mohan 99

Infinitesimal Generator

Prof. Rajasekar Mohan 100

Infinitesimal Generator

Prof. Rajasekar Mohan 101

State Transition Diagram M/M/1

Prof. Rajasekar Mohan 102

Queue Performance Measures

Prof. Rajasekar Mohan 103

Queue Performance Measures

Prof. Rajasekar Mohan 104

Multiplexing

Consider the M/M/1 queueing model where traffic from multiple sources
that follow Poisson processes enters an infinite length buffer and are
served by a single server with exponential service time. This model
enables us to gain insight into efficiency gain of multiplexing

Prof. Rajasekar Mohan 113

Multiplexing

Assume that our traffic model obeys the M/M/1 assumptions. Then if the
arrival rate increases from λ to Nλ and we increase the service rate from μ
to Nμ (maintaining the same ρ), the mean queue-size and its distribution
will remain the same. However, in this scenario the mean delay does not
remain the same. It reduces by N times to 1/[N(μ − λ)].

This is applicable to a situation where we have N individual M/M/1 queues each of

which with arrival rate λ and service rate μ. Then we superpose (multiplex) all the
arrival processes together which results in a Poisson process of rate Nλ. If we replace
all the individual servers (each of which has service rate μ) with one fast server
that serves the superposed Poisson stream of rate Nλ, what service rate this fast
server should operate at ?.

Prof. Rajasekar Mohan 114

Multiplexing
If our QoS measure of interest is the mean delay, or the probability that
the delay exceeds a certain value, and if for a given arrival rate λ there is
a service rate μ such that our delay-related QoS measure is just met, then
if the arrival rate increases from λ to Nλ, and we aim to find the service
rate μ∗ such that the delay-related QoS measure is just met, we will need
to make sure that the spare capacity is maintained, that is
μ − λ = μ∗ − Nλ or μ∗ = μ + (N − 1)λ
so by the latter and the stability condition of μ > λ, which implies
that the μ + (N − 1)λ is less than μ + (N − 1)μ for N > 1, we
must have that μ∗ < Nμ for N > 1.

Prof. Rajasekar Mohan 115

Multiplexing

Let us define a measure for multiplexing gain to be given by

Prof. Rajasekar Mohan 116

Multiplexing

Prof. Rajasekar Mohan 117

Multiplexing Gain

• If ρ is very close to 1, then the multiplexing gain diminishes – queues are already efficient.

• If QoS requirements are strict (requiring very low mean queueing delay), then utilization ρ
is low, then the potential for multiplexing gain is high.

Prof. Rajasekar Mohan 118

Multiplexing Gain
TDMA - each user is assigned one or more channels (in a form of time-
slots) to access the network.

Full multiplexing FMUX – users send data to a switch which then

forwards the data to the destination. That is, all the data is stored in one
buffer (in the switch) which is served by the entire available link
capacity.

Prof. Rajasekar Mohan 119

Multiplexing Gain

Prof. Rajasekar Mohan 120

Multiplexing Gain

Prof. Rajasekar Mohan 121

M/M/1 model – Achieving desired QoS
(contd.)

Suppose the mean delay E[D] < ε or rather P(D > t) <
α, where α is a percentile of the delay distribution, is
to be achieved.
Suppose α and λ are given then, minimal service rate
is chosen as follows.

Prof. Rajasekar Mohan 133

M/M/1 model – Achieving desired QoS
(contd.)
Suppose α and μ are given and the delay constraint must
be satisfied then, maximal arrival rate must be chosen as
follows.

This way the stability λ < μ will not be violated and a

feasible λ can be found
Prof. Rajasekar Mohan 134
M/M/1 model – Example

Variable-length data packets arrive at a communication

node according to a Poisson process with an average
rate of 180 packets per minute. The single outgoing link
is operating at a transmission rate of 4800 bps. The
packet length can be assumed to be exponentially
distributed with an average length of 120 characters in 8
(7+1) ASCII format. Calculate the principal performance
measures of this communication link assuming that it
has a very large buffer. What is the probability that 7 or
more packets are waiting to be transmitted?

Prof. Rajasekar Mohan 135

M/M/1 model
(contd.)

Solution:

Prof. Rajasekar Mohan 136

M/M/1 model – Example

In an ATM network, two types of packets arrive at a single-

channel transmission link; that is voice packets that are
always accepted and data packets that are only accepted
when the total number of packets in the system is less
than N>0. Find the steady-state probability mass function
of having k packets in the system if both packet streams
are Poisson with rate λ1 and λ2 respectively. All packets
have exponentially distributed lengths and are
transmitted at an average rate μ.

Prof. Rajasekar Mohan 137

M/M/1 model – Example
Solution:

Prof. Rajasekar Mohan 138

M/M/1 model – Example
(contd.)

Prof. Rajasekar Mohan 139

Offered & Carried Traffic

Prof. Rajasekar Mohan 140

 Offered & Carried Traffic
• The offered traffic intensity and carried traffic intensity are synonymous
to offered traffic and carried traffic, respectively.

• In cases of infinite buffer/capacity systems, such as M/M/1 and M/M/∞,

the term traffic intensity is used to describe both offered traffic and
carried traffic.

• Accordingly, ρ is called traffic intensity in the M/M/1 context and A is

the traffic intensity in an M/M/∞ system. Others use the term traffic
intensity in multiservice system for the offered load per server.

• To avoid confusion, the term traffic intensity in the context of a single

server queue with infinite buffer is used, in which case, traffic intensity
is always equal to ρ = λ/μ.
Prof. Rajasekar Mohan 141
Offered & Carried Traffic

Prof. Rajasekar Mohan 142

Offered & Carried Traffic

Prof. Rajasekar Mohan 143

Multi-server queuing model

m servers in parallel with service rate μ

• Service times follow exponential distribution

Poisson arrivals with rate λ

Queuing discipline is FIFO

Prof. Rajasekar Mohan 147

Multi-server queuing model
Two models are considered:
a) infinite servers and
b) finite servers

Offered traffic: It is defined as the mean number of

arrivals per mean service time (i.e., λ/μ) .
• It is denoted by A and expressed in erlangs
• A can be greater than 1 while 0 < ρ < 1 in M/M/1
• Both A in M/M/ꚙ model and ρ in M/M/1 represent
the mean number of busy servers in the system

Prof. Rajasekar Mohan 148

Multi-server queuing model

Carried traffic: It is defined as the mean number of

customers, calls or packets leaving the system after
completing service during a time period equal to the mean
service time.

• Also measured in erlangs

• Offered traffic versus carried traffic
• For M/M/ꚙ system the carried traffic is equal to
offered traffic.
• In more realistic M/M/k system, offered traffic is always

higher than carried traffic due to congestion

Prof. Rajasekar Mohan 149
M/M/ꚙ model

As there are infinite number of servers each with

service rate μ, the mean time spent by an arrival in the
system is 1/μ.
Therefore, by Little’s theorem, the mean capacity of
the system is given by

Carried traffic is equal to the mean number of busy

servers.
It is also equal to the mean queue size because carried
traffic is equal to offered traffic

Prof. Rajasekar Mohan 150

M/M/ꚙ model – State transitions
As in M/M/1 model the queue size represents the
state of the system
Arrival rate is independent of the queue size
However, service rate changes with queue size
• When there are n customers in the system, and at
the same time, n servers are busy, the service rate
is nμ.
The time till the process stays in any state is
exponentially distributed with λ+nμ
• This is because there is a competition between n
exponential random variables (with parameter μ)
and one exponential random variable (with
parameter λ)
Prof. Rajasekar Mohan 151
M/M/ꚙ model – State transitions

Prof. Rajasekar Mohan 152

M/M/ꚙ model – State transitions

Prof. Rajasekar Mohan 153

Steady-State Equations

We observe that the distribution of the number

of busy channels (simultaneous calls or
customers) in an M/M/∞ system is Poisson with
parameter A.

Prof. Rajasekar Mohan 154

M/M/ꚙ model – State transitions
(contd.)
Let πi denote the state probability of having i ≥ 0
customers/packets in the system

Using local balance concept

Prof. Rajasekar Mohan 155

M/M/ꚙ model – State transitions
(contd.)

By induction we get

The infinitesimal generator matrix Q is given by

Prof. Rajasekar Mohan 156

M/M/ꚙ model – State transitions
(contd.)
Using the recursive local balance equation

and the normalization equation, we can determine

the steady state distribution

For M/M/ꚙ model the state distribution is a Poisson

process of busy servers
Prof. Rajasekar Mohan 157
M/M/m/ꚙ model

Using local balance concept for n ≤ m

Prof. Rajasekar Mohan 158

M/M/m/ꚙ model
(contd.)
Using local balance concept for n > m

Using normalization equation, we get

Prof. Rajasekar Mohan 159

M/M/m/ꚙ model
(contd.)

The probability that a new arrival finds all servers are

busy and hence has to wait is given by

The above equation is called Erlang C formula

Prof. Rajasekar Mohan 160

M/M/m/ꚙ model
(contd.)
Queue forms only when all servers are busy
Mean queue size E[Nq] is given by

Prof. Rajasekar Mohan 161

M/M/m/ꚙ model
(contd.)
Mean waiting time in the queue is obtained by
Little’s theorem

Mean time spent by a customer/packet in the

system is given by

Prof. Rajasekar Mohan 162

 Assignment
1. M/M/k/k QUEUEING SYSTEM (Reference : Chapter 8 - Queueing Theory and Stochastic
Teletraffic Models)

2. Review of One Journal Paper (Later than or of 2011) related to Wireless Networking
Format for review :
-https://drive.google.com/file/d/1n6T4iXmyXAgD3e0_TiHnna4Ti4stcfCX/view?usp=sharing

3. Simulation /Hardware – (Group of 2 students Max) – Topics to be Informed

Instructions :

• Individual submission with SRN /Sec on Top Right of every sheet

• No. of Pages : At least 7 pages
• Soft copy of the Journal reviewed to be shared
• Simulation / Hardware assignments will be demonstrated
• All submissions (hand-written) can be in softcopy /hard copy
• Deadline for finalization of marks : 20 Nov 2020 – Sl No.1 & 2 ; 25 Nov 20 – Sl No.3
Prof. Rajasekar Mohan 177