Network architecture

• INTRODUCTION
 Networks Matrix
• Fixed versus mobile: fixed mobile Convergence with common IP

• General public (box @, low cost solution) versus corporate

(advanced features: security , Qos, … )

• Voice versus data (voice data convergence same physical

network (abandon du RTC en 2023 en France, ARCEP OK). It
involves new and additionnal security issues.
 Data and voice Network
3 types of fixed networks: LAN , WAN, local
loop.

LAN
WAN

local loop

ROUTER
Transmission network

PABX
 Network topology
POINT-to-POINT

Local loop N1 N2

STAR
N1
N2 N3

N2 N3
N1 LAN

N4 N5
N4 N5 N6
MULTI-POINT: bus ring … not used any more

N2
N1 N3

N1 N2 N3 Nx

N4 Nx
N5
 Meshed networks

N1

WAN (redundancy)
N2 N3

N4 N5 N6
 Network security
• Basic rules:
• 1. Network security is part of network architecture and
not an add-on. Network security must be integrated at
the very beginning of any new Network project. After it is
too late (heavy and/or uneffective) . Ex. IPSEC & NAT
• 2. Network security requires an indepth knowledge about
network architecture (Only a network expert can become
a network security expert) …. General introduction
before network security part
• 3. Security (against attacks but also unintentional
accident … example) and Reliability (resiliency) are 2
different points to deal with: no confusion (french
:sécurité/ sureté)
 Information Natures

VOICE
 Analog signal
 300 - 3400 Hz
 Hi-Fi: up to 30000 Hz
 Can be converted in digital
DATA
 digital
 Code required
VIDEO
VOICE
Can accept transmission errors
very sensible to jitter and delay problems …
DATA
not sensible to delays
Very sensible to errors

VIDEO similar to voice

Before , voice transport have dedicated networks (PSTN e.g.),

Today: voice and data convergence : same physical network for
voice and data: security issue for voice channel
Consequence: new issues : IP-PBX attacks (intrusions permits
charged calls generation: undetected, very remunerative « job » )
Period, frequency and wavelength

Imaginons un voyageur jouant au yo-yo dans un

TGV lancé à 360 km/h (100 mètres par seconde).
La bobine du yo-yo met deux secondes pour
descendre puis remonter à son point de départ,
on peut dire que la période de son mouvement
est égale à 2 secondes. Si T = 2s on peut calculer
facilement la fréquence du mouvement :
En 2 secondes le train, le voyageur et son
yo-yo se sont déplacés de 200 m et la
bobine du yo-yo a eu le temps de dessiner
une courbe correspondant à une période
entière.

La distance de 200 m correspond à la

longueur d'onde du mouvement
oscillatoire du yo-yo.
La figure 1 représente le mouvement du yo-yo dans le
temps (avec la période T) et la figure 2 ce mouvement
dans l'espace (avec la longueur d'onde l ).

La longueur d'onde dépend non seulement de la

fréquence du mouvement mais aussi de la vitesse de
déplacement du train, on peut résumer cette relation à
l'aide de la formule :
 Wavelength and security ??

• Let’s take the example of local loop with optical fiber

• With passive optical network , and for cost reasons one

fiber is installed for 32/64 customers (with passive splitter)
• So all the 32/64 customers receive the same information at
home: security weakness
• If one dedicated wavelength is assigned to each customer:
no risk any more
• Expensive solution: the best compromise security/cost
must be found …
 Definitions

Bandwidth (Hz) :
BW = F maximal - F minimal

Often confusion between bandwidth and rate ….

Rate is b/s
 Definitions
Rate (bits/s ou bps) :

R = BW log2 (1 + S/N) S/N = signal/noise ratio*

.
BW ≈ 3400 Hz
S/N ≈ 1000 => Rate≈ 34000 bits/s

log2 10 = 3,32.
Log2 2P10 = 10
 Sine Wave

+
1/2 Cycle
0 TIME
1/2 Cycle
-

1 Cycle

Digital Signal

0 TIME

-
 Telephony Voice Bandwidth
(or, your dial-up modem operations)

Energy: Most of the voice signal’s energy is

carried in the vowel sounds which
reside in the 250 - 500 Hertz area

Articulation:
Understanding, provided by
consonants, resides in the areas
between 700 - 3000 Hertz

This lead to the voice band being defined as

the frequency range between 200 and 3200 Hertz
 Telephony Voice Band

Vowels Consonants

ARTICULATION
POWER

200 500 700 3000 3200

Hertz
 Advantages and Disadvantages of Analog

• Natural - Easily Understood and Handled

• Adequate Bandwidth for Voice
• Traditionally, Lower Equipment Costs

• Noise Amplified and Forwarded

• Not “data” Friendly
• Limited Cross Connect Capabilities
Advantages and Disadvantages of Analog

Soup Nuts
Base Band: No Modulation (BdB)
No frequency transposition

Avantages :
Low cost

Drawbacks: attenuation … only for short distances

 Advantages and Disadvantages of Digital

•Digital Signal Regeneration “blocks” noise

• Performance Monitoring and Error Correction
• All messages are 0s and 1s

• Cost Incurred for Analog to Digital to Analog

Conversion
• Need Synchronization - A Common Clock
• Increased Bandwidth Requirements
 Advantages and Disadvantages of Digital

“Clock” Needed to Identify Start

Digital Transmission Systems Require up to 16

time the bandwidth of an analog system.

Costs for A/D/A Conversion

 Analog to Digital Conversion

•The information carried by the continuously varying analog

signal must be manipulated:

• The analog signal has an infinite range of possibilities

• The digital signal has a finite range of possibilities, 1 and 0

• Pulse Code Modulation is the technique that encodes the

analog signal into an 8 bit word

• This is done by encoding an “instantaneous” value of the

analog signal.
• This “instantaneous” value is rebuilt into the original
analog waveform at the receiving end
 PCM Transmission

Terminating
Originating
Office
Office
1 0 1 0 1 0 1 1

PCM PCM
Encoder Decoder

8 bit word represents

the analog signal at
a single point in time.
 Binary Encoding and Decoding

numbers are 0 to 9; from these, we can count to any number.It is

referred to as the Base 10 System.

The Binary Numbering System consists of two digits, 0 and 1.

Those are the only two conditions. This Base 2 System can
also represent all possible numbers. However, it does take a few
more characters to represent numbers beyond 0 and 1.
BINARY DECIMAL

0 0
1 1
10 2
11 3
100 4
101 5
110 6
111 7
1000 8
1001 9
1010 10
 How to Convert Binary to and from Base 10

POWER 27 26 25 24 23 22 21 20

WEIGHT 128 64 32 16 8 4 2 1

1 0 1 0 0 0 0 0 = 161

0 1 1 1 0 0 1 0 = 114

1 1 0 1 1 0 1 1 = 219

0 0 1 0 1 1 0 1 = 45

1 1 1 1 1 1 1 1 = 255
 The Binary Number in Digital Communications

POWER 27 26 25 24 23 22 21 20

WEIGHT +/- 64 32 16 8 4 2 1

BIT POSITION 1 2 3 4 5 6 7 8
Practice

- 37 = 0 0 1 0 0 1 0 1

+15 = 1 0 0 0 1 1 1 1

-0= 0 0 0 0 0 0 0 0

+101 = 1 1 1 0 0 1 0 1
 Analog to Digital Conversion Process

Converting an analog signal to a PCM (digital) Signal is a three step process

• Sample
the analog signal at a rate that is twice its frequency; this rate is
called the Nyquist Frequency. This results in a Pulse Amplitude
Modulation (PAM) signal which has an amplitude equal to the amplitude
of the sampled analog signal at the moment of sampling.
• Quantize
the PAM signal into one of 256 possible levels. This is done using
either the -255 (mu-255) (North American Standard) or the A-Law
(CEPT Standard) scales to derive a NUMBER that represents the PAM
signal’s amplitude.
• Encode
the number determined by the scale into a binary number. This
process has you converting a base 10 number to a base 2 number.
The final result is an 8-bit binary number.
 Analog to Digital Conversion Process

1 1 1 0 0 1 0 0

Sample Quantize Encode

PAM
Signal PCM Signal

+127

+100 Base 10 100

0 Base 2 ?

-127

-255 scale
 Sampling

Our sample rate is 8000 times per second.

This rates comes by applying the Nyquist
Theorem.
This theorem states:

When a band limited signal that is sampled at

regular intervals and at a rate greater than or
equal to twice the highest frequency, then the
samples contain all the information of the
original signal.
Our telephony voice signal is between 200 - 3200
Hertz.
If we “assumed” a total bandwidth of 4000 Hz,
then the sampling rate would be (2 X 4000 =) 8000
samples per second.
Thus, we have more than enough samples to
carry ALL the information of the original analog
signal.
Every second the system obtains 8000 PAM
samples; the PAM sample is the amplitude of the
analog signal at an instant in time.
 Sampling

TIME
Amplitude

125 sec
1/8000 sec
125 sec
1/8000 sec

PCM Bits per second being transmitted: ?

 Quantizing
After the analog signal has been sampled and a PAM
signal obtained, thenext operation is to “quantize” the
PAM signal. This is done by comparing the PAM
signal to a scale, such as the -255, to determine the
value of the amplitude of the signal. This will be an
integer between -127 and +127.
This is the information (the amplitude of the PAM
signal) that is passed to the encoder to create the
PCM signal.
The output to the encoder is an integer between -127
and +127.
Therefore, if the PAM signal does not exactly
match a value, the resulting value being sent to
the encoder must be rounded up or down to a
whole number.

This introduces Quantizing Error or Distortion.

The PCM signal created from this information is
an integer. Therefore, at the receiving end, the
PAM signal will also be an integer.The PAM signal
produced will be a whole number, and the
resulting recreated analog signal will be just a
little different from the originating
signal...distortion is part of the process.
 Quantizing

+ 127
PAM Signals

100
Amplitude

76 TIME
125 sec

1/8000 sec
125 sec
0
1/8000 sec

TIME
-
89
Quantizing Error
102 or

127 Distortion

- 89
- 90
 Encoding

Encoding is the conversion of the quantized PAM

Signal, which is in Base 10, to a Base 2, or Binary,
number. Since our PAM Signal can be any whole
number between -127 and +127 (this equates to 256
8
possible choices) we need an 8 bit word (2 = 256).
This is something we have already done. The first
position on the far left
determines if the number is +/-. The next seven
positions is the number.
This number is then transmitted as a PCM signal; it
is transmitted as a serial set of 0s and 1s.
 Encoding

0 1 0 11 0 0 1
TIME
Encoder

Quantizing Error P P
or A A
M M PCM
Distortion Signal
or
- 89 Binary
- 90 Quantized
Amplitude Word
is or
- 89 Time
Slot
 Decoding
To go from Digital back to Analog requires a
decoder to convert the PCM Signal, which is in
Binary form, to a PAM Signal, which is in Base 10.
Thus, each PCM word will be converted to the PAM
amplitude that it represents.

Once the PAM Signal is recreated, a filter is used to

reconstruct the original analog signal.

The generic name of the equipment used for this

work is a CODEC, which stands for Coder-Decoder.
 Modulation of a signal
S(t) = A sin (2 F t + )
 Modulation Effects

Noise inversely proportional to frequency

Short distance: no need for modulation: baseband – ethernet

LAN

Long distance: modulation needed (local loop up to 5 kms)

That is why internet access is also called broadband
(modulation) ADSL is an example
 Different types of multiplexing
Fixed: simple but wasted ressources
1 2 3 4 5 6 7 8
Statistical (overbooking ratio to define: complex or
contention )
1 2 3 4 5 6 7 8

With or without compression

 Frequency Multiplexing

AAAAAA AAAAAA F1
BBBBBBB BBBBBBB F2
CCCCCCC CCCCCCC F3
TDM (Time Division Multiplexing)

AAAAAA
BBBBBBB CC BB AA CC BB AA CC
CCCCCCC
t

1 Time slot per channel (IT)

 T1 - TDM - Time Division Multiplexing

TDM operates on the fact that the samples are taken every
125 sec (8000 samples per second) AND the time to take a
sample is significantly less than 125 sec.
Thus, there is time to take additional samples.

These samples of other channels are taken during the

“idle” time in the 125 sec available between samples of
the first channel.
When Time Division Multiplexing was applied in the early
1960s, it was possible to take a total of 24 different
samples in the 125 sec.
These 24 samples were 24 PAM pulses which were then
quantizied and encoded into 8-bit PCM signals.

Since each channel was sampled 8000 times per second, and
each sample resulted in an 8 bit PCM signal, each voice
grade channel was the equivalent of 64,000 bits/sec.
24 such channels equates to 1,536,000 bits per second.

With one frame bit per 24 samples, we have a signal of

1,544,000 bits per second OR 1.544 Mb/s, which is a T1
signal.
 T1 - TDM - Time Division Multiplexing

1/8000 sec
SAMPLE
Channel 1

SAMPLE 1/8000 sec

Channel 2

1/8000 sec
Chan 1
SAMPLE 1/8000 sec Chan 2
Channel 3

Chan 3
Quantizer

Encoder Time Time

Slot
Time
Slot Time
Slot 3 2
SAMPLE 1/8000 sec 24 Slot
Channel 24
1
Chan 24

En Europe : E1: 2048 kb/s soit 32 IT

ISDN Integrated Services Digital Network

* BRI Basic Rate Interface

BRI = 2B+D
144kbs = [64kbs] + [64kbs] + (16kbs)

* PRI Primary Rate Interface

PRI = 23B + D
T1 = 23[64kbs] + {64kbs}

En Europe : accès primaire contient 30 canaux B

 Commutation: Switching

 circuit switching
 packet switching
 Circuit switching

N1

N2 N3
information

N4 N5 N6

fixed delay no jitter

ideal for voice but wasted ressources (one person speaks 35 % of the time)
PSTN , POTS networks
francele réseau téléphonique commuté (RTC)
 packet switching

N1 information

N2 N3
information

N4 N5 N6

Virtual circuit
ressources optimization
Variable delay, non deterministic
More complex protocols: QoS required if limited BW

 X25, ATM, Frame Relay,, …

 CLIENT: TERMINAL
LL POP WAN
PC or PAD
N CV3
switch X25
CV1 N N

N
N N
CV2
VC1+2+3
CV1 CV2 CV3

SERVER
 Connexion modes

CONNECTIONLESS

  LANs (free bandwidth, broadcast). No needed configuration

CONNECTION ORIENTED

 For WAN
pre-defined VC required (ATM, X25, FR, …)
 Connexion modes

Reliable:
Controles (flow, integrity, …)
Example: TCP, X25

Unreliable: better performance

No controle
Example: UDP
 Layers model OVERVIEW

• THE NEED FOR STANDARDS

• OSI - ORGANISATION FOR
STANDARDISATION
• THE OSI REFERENCE MODEL
• A LAYERED NETWORK MODEL
• THE SEVEN OSI REFERENCE MODEL
LAYERS
• SUMMARY
 THE NEED FOR STANDARDS

• Over the past couple of decades many of the networks that were
built used different hardware and software implementations, as a
result they were incompatible and it became difficult for networks
using different specifications to communicate with each other.
• To address the problem of networks being incompatible and
unable to communicate with each other, the International
Organisation for Standardisation (ISO) researched various
network schemes.
• The ISO recognised there was a need to create a NETWORK
MODEL that would help vendors create interoperable network
implementations.
 ISO - ORGANISATION FOR
STANDARDISATION

• The International Standard Organisation (ISO) is an

International standards organisation responsible for a
wide range of standards, including many that are
relevant to networking.
• In 1984 in order to aid network interconnection without
necessarily requiring complete redesign, the Open
Systems Interconnection (OSI) reference model was
approved as an international standard for
communications architecture.
 THE OSI REFERENCE MODEL
• The model was developed by ISO in 1984. It is now considered
the primary Architectural model for inter-computer
communications.
• The Open Systems Interconnection (OSI) reference model is a
descriptive network scheme. It ensures greater compatibility and
interoperability between various types of network technologies.
• The OSI model describes how information or data makes its way
from application programmes (such as spreadsheets) through a
network medium (such as wire) to another application programme
located on another network.
• The OSI reference model divides the problem of moving
information between computers over a network medium into
SEVEN smaller and more manageable problems .
• This separation into smaller more manageable functions is known
as layering.
 A LAYERED NETWORK MODEL
• The process of breaking up the functions or tasks of
networking into layers reduces complexity.
• Each layer provides a service to the layer above it in the
protocol specification.
• The lower 4 layers (transport, network, data link and
physical —Layers 4, 3, 2, and 1) are concerned with the
flow of data from end to end through the network.
• The upper four layers of the OSI model (application,
presentation and session—Layers 7, 6 and 5) are
orientated more toward services to the applications.
• Data is Encapsulated with the necessary protocol
information as it moves down the layers before network
transit.
THE SEVEN OSI REFERENCE MODEL
LAYERS
 OSI model: 7 layers

7: APPLICATION •APPLICATION LIBRARY

6: PRESENTATION •DATA FORMAT
•INTER APPLICATION
5: SESSION
MANAGEMENT (Synchro)
4: TRANSPORT •END TO END CONTROL
•ROUTING INFORMATION
3:NETWORK
THROUGH NETWORK
2:DATA LINK •ACCESS METHOD, FRAME
FORMAT
1: PHYSICAL
•MEDIA ACCESS
 7 APPLICATION APPLICATION 7

PRESENTATION
PROTOCOLS PRESENTATION
6 6

MESSAGES
5 SESSION SESSION 5

4 TRANSPORT TRANSPORT 4

3 NETWORK PACKETS NETWORK 3

2 DATA LINK FRAMES DATA LINK 2

SERVICE
1 PHYSICAL BITS PHYSICAL 1
PHYSICAL MEDIA
 LAYER 7: APPLICATION

• The application layer is the OSI layer that is closest to

the user.
• It provides network services to the user’s applications.
• It differs from the other layers in that it does not
provide services to any other OSI layer, but rather,
only to applications outside the OSI model.
• The application layer establishes the availability of
intended communication partners, synchronizes and
establishes agreement on procedures for error
recovery and control of data integrity.
 LAYER 6: PRESENTATION

• The presentation layer ensures that the

information that the application layer of one
system sends out is readable by the
application layer of another system.
• If necessary, the presentation layer translates
between multiple data formats by using a
common format.
• Provides encryption and compression of data.
 LAYER 5: SESSION

• The session layer defines how to start, control and end

conversations (called sessions) between applications.
• This includes the control and management of multiple bi-
directional messages using dialogue control.
• It also synchronizes dialogue between two hosts'
presentation layers and manages their data exchange.
• The session layer offers provisions for efficient data
transfer.
 LAYER 4: TRANSPORT
• The transport layer regulates information flow to ensure end-to-end
connectivity between host applications reliably and accurately: for
instance flow control, multiplexing
• The transport layer segments data from the sending host's system
and reassembles the data into a data stream on the receiving host's
system.
• The boundary between the transport layer and the session layer
can be thought of as the boundary between application protocols
and data-flow protocols.
• Layer 4 protocols is similar with TCP (Transmission Control
Protocol) and UDP (User Datagram Protocol) in the TCPIP model.
Similar but not compatible
 LAYER 3: NETWORK

• Defines end-to-end delivery of packets.

• Defines logical addressing so that any endpoint can
be identified.
• Defines how routing works and how routes are
learned so that the packets can be delivered.
• The network layer also defines how to fragment a
packet into smaller packets to accommodate different
media.
• Routers operate at Layer 3.
• Similar with IP protocol in TCPIP model
 LAYER 2: DATA LINK
• The data link layer provides access to the networking media and
physical transmission across the media and this enables the data to
locate its intended destination on a network.
• The data link layer provides reliable transit of data across a physical
link by using the Media Access Control (MAC) addresses.
• The data link layer uses the MAC address to define a hardware or
data link address in order for multiple stations to share the same
medium and still uniquely identify each other.
• Concerned with network topology, network access, error
notification, ordered delivery of frames.
• Examples :8802.3, Frame Relay, FDDI.
• 8802.3 is very similar with Ethernet (linked withTCP/IP model) but
… still no compatibility … only cohabitation … fortunately
 LAYER 1: PHYSICAL
• The physical layer deals with the physical characteristics
of the transmission medium.
• It defines the electrical, mechanical, procedural, and
functional specifications for activating, maintaining, and
deactivating the physical link between end systems.
• Such characteristics as voltage levels, timing of voltage
changes, physical data rates, maximum transmission
distances, physical connectors, and other similar attributes
are defined by physical layer specifications.
• Examples :- EIA/TIA-232, RJ45, NRZ, codes
 Code features

• Average = 0 (even if many successive 1 or 0)

• Auto-synchronization (no need to transmit clock separately)

• Possible Error detection and correction ( with redundant

bits)
 SUMMARY
• There was no standard for networks in the early days and as a
result it was difficult for networks to communicate with each other.
• The OSI reference model has standards which should ensure
vendors greater compatibility and interoperability between various
types of network technologies.
• The OSI reference model organizes network functions into seven
numbered layers.
• Each layer provides a service to the layer above it in the protocol
specification and communicates with the same layer’s software or
hardware on other computers.
• Layers 1-4 are concerned with the flow of data from end to end
through the network and Layers 5-7 are concerned with services to
the applications.
 But ….
• TCP/IP is the one we use now

• Layer 5, 6 and 7 are merged in the application

itself (layer « 8 »)

• OSI model was too slow too complex … and at

the end do not permit sufficient interoperability

• Remember that TCP/IP and Ethernet are not

compatible with OSI layer 3 and 4 and 8802.3
TCPIP, Ethernet : successful with
good performance cause easy
protocols .

Easiness means security flaws

Examples : ARP, DHCP, DNS,

switches with auto-learning table,
SNMP, …
Equipment A Equipment B

7: APPLICATION 7: APPLICATION

6: PRESENTATION
GATEWAY 6: PRESENTATION

5: SESSION 5: SESSION

4: TRANSPORT 4: TRANSPORT

3:NETWORK 3: ROUTER 3: NETWORK

2: DATA LINK 2: BRIDGE/SWITCH 2: DATA LINK

1: PHYSICAL 1: REPEATER 1: PHYSICAL

Network 1 Network 2