Data Communication Fundamentals

1

2010/02/15(C) Herbert Haas

Communication Basics

Principles and Dogmas

In this chapter we discuss basic communication issues, such as synchronization,

coding, scrambling, modulation, and so on.

2

Everything

should be made

as simple as possible,

...but not simpler.

Albert Einstein

3

3(C) Herbert Haas 2010/02/15

Information

What is information?Carried by symbols

Recognized by receiver (hopefully)

Interpretation is the key

What is information? This question may sound quite easy but think a bit about it.

Obviously we need symbols to represent information. But these symbols must

also be recognized as symbols by the receiver. In fact, philosphical considerations

conclude that information can only be defined through a receiver. The same

problem is with art. What is art? Several decades and centuries had their own

definitions. Today most critics use a general definition: art can only be defined in

context with the viewer.

In the following chapters throughout the whole data communication we will

deal with symbols representing information. A symbol is not a 0 or a 1. But this

binary information can be represented by symbols. Be patient...

4


Symbols

Symbols (may) represent information

Voice patterns (Speech)

Sign language, Pictograms

Scripture

Voltage levels

Light pulses

Blue Whale Sonagrams

What is a good information source? From a theoretical point of view a random

pattern is the best because you'll never know what comes next. On the other hand,

if you receive a continous stream of the same symbol this would be boring. More

than boring: there is no information in it, because you can predict what comes

next! From this we conclude that a sophisticated coding representing the

information as efficient as possible using symbols is a critical step during the

communication process.

Throughout these chapters we will mainly deal with symbols such as voltage

levels or light pulses.

Look at the Blue Whale Sonograms. The x-axis represents time, the y-axis

frequency and the color represents power density. This communication pattern is

very complex (those of dolphins is even more complex). It is known that each

herd has their own traditional hymn. And: they like to communicate!

5


Symbols on Wire

Discrete voltage levels = "Digital"

Resistant against noise

How many levels?

Binary (easiest)

M-ary: More information per time unit!

Binary M-ary

(here 4 levels, e. g. ISDN)

What symbols do we encounter on wire? Digital binary symbols are commonly

known and widely in use. Why? Consider information transmissions in groups of

symbols (for example the group of 8 binary symbols is called a byte). We have

two parameters: the number base B and the group order C. If you calculate the

"costs" that you get for arbitrary variations of B and C, and if we assume a linear

progress (so that cost =kBC) then for any given (constant) cost the perfect base

would be B=e, that is B=2.7182...

In other words: the perfect base is a number between 2 and 3. The technical

easiest solution is to use B=2. Note that these considerations assume a linear cost

progression.

In many cases we pay the price of higher efforts and use a larger base. This leads

us to m-ary symbols and later to PAM and QAM.

6


Synchronization

Sender sends symbol after symbol...

When should receiver pick the signal samples? => Receiver must sync with sender's clock !

?00001

00001100110

000100111111

001010010111

Sampling instances Interpretation:

(only this one is correct)

One of the most important issues among communication is that of

synchronization. Nature forbids absolute synchronization of clocks. Suppose you

are a receiver and you see alternating voltage levels on your receiving interface. If

you had no idea about the sending clock then you would never be able to

interprete the symbols correctly. When do you make a sample?

7


Synchronization

In reality, two independent clocks are NEVER precisely synchronous We always have a frequency shift

But we must also care for phase shifts

?

001010011110

????????????

001010011011

Phase shift

(worst case)Different

clock

frequencies

So we must assume that the receivers clock is approximately identical to the

senders clock. At least we must deal with small phase and frequency gaps. As you

can see in the slide above, we still cannot be sure when to make samples.

What we need is some kind of synchronization method.

8


Serial vs Parallel

Parallel transmission Multiple data wires (fast)

Explicit clocking wire

Simple Synchronization but not cost-effective

Only useful for small distances

Serial transmission Only one wire (-pair)

No clocking wire

Most important for data communication

In case of parallel transmissions there is always a dedicated clock line. This is a

very comfortable synchronization method. A symbol pattern on the data lines

should be sampled by the receiver each time a clock pulse is observed on the

clock line. But unfortunately, parallel transmissions are too costly on long links.

In LAN and WAN data communication there are practically no parallel lines.

The most important transmission technique is the serial. Data is transmitted over

a single fiber or wire-pair (or electromagentic wave). There is no clock line. How

do we synchronize sender and receiver?

9


Asynchronous Transmission

Independent clocks Oversampling: Much faster than bitrate

Only phase is synchronized Using Start-bits and Stop-bits

Variable intervals between characters

Synchronity only during transmission

Inefficient

Character Character Character

Stop-Bits

Start-

Edge

Start-Bit

Variable

One synchronization method is the Asynchronous Transmission. Actually this

method cannot provide real synchronization (hence the name) but at least a short-

time quasi-synchronization is possible. The idea is to frame data symbols using

start and stop symbols (lets sloppy call them start- and stop bits). Using

oversampling, the receiver is able to get a sample approximately in the middle of

each bit but only for short bit-sequences.

Asynchronous transmission is typically found in older character-oriented

technologies.

Example application: RS-232C

Relative overhead: 3/11

10


Synchronous Transmission

Synchronized clocks Most important today!

Phase and Frequency synchronized

Receiver uses a Phased Locked Loop (PLL) control circuit Requires frequent signal changes

=> Coding or Scrambling of data necessary to avoid long sequences without signal changes

Continous data stream possible Large frames possible (theoretically endless)

Receiver remains synchronized

Typically each frame starts with a short "training sequence" aka "preamble" (e. g. 64 bits)

The most important method is the Synchronous Transmission. Don't confuse this

with synchronous multiplexingwe are still on the physical layer! Two things are

necessary: a control circuit called Phased-Locked-Loop (PLL) and a signal that

consists of frequent transitions. How do we ensure frequent transitions in our data

stream? Two possibilities: coding and scrambling our data.

Synchronous Transmission is found in most modern bit-oriented technologies

nearly anything you know.

11


Line Coding

1 0 1 0 1 1 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0NRZ

RZ

Manchester

DifferentialManchester

NRZI

AMI

HDB3

Code Violation

The trivial code is Non Return to Zero (NRZ) which is usually the human naive

approach.

Remarks:

RZ codes might also use a negative level for logical zeroes, a positive level for

logical ones and a zero Volt level inbetween to return to. RZ is for example used

in optical transmissions (simple modulation).

NRZI codes either modulate for logical ones or zeros. In this slide we modulate

the zeroes, that is each logical zero requires a transition at the beginning of the

interval.

NRZI means Non-return to zero inverted or interchanged.

B8ZS: same as bipolar AMI, except that any string of eight zeros is replaced by a

string of two code violations.

Manchester is used with 10 Mbit Ethernet. Token Ring utilizes Differential

Manchester. Telco backbones (PDH technology) use AMI (USA) or HDB3

(Europe). Of course there are many many other coding styles.

12


Power Spectrum Density

0.5 1.51.0 2.0

1.0

0.5

NRZ,

NRZI

HDB3

AMI

Manchester,

Differential Manchester

Normalized

Frequency (f/R)

Spectral

Density

The slide above compares the power density distribution of some codes

mentioned before. Obvously the code must match the spectral characteristics of

the transmission channel.

Note that these codes are still kinds of baseband transmissions. Each one can be

modulated using a carrier signal at higher frequency to comply to a specific

channel characteristic.

13


Scrambling Example

TS

TS

TS

TS

TS

TS

TS

TS

TS

TS

TS

TS

TS

TSChannel

Example:

Feedback Polynomial = 1+x4+x7

Period length = 127 bit

t(n-4) t(n-4)

t(n-7) t(n-7)

s(n) t(n) t(n) s(n)

Another method to guarantee frequent transitions is scrambling. Scramblers are

used with ATM, SONET/SDH for example.

The feedback polynomial above can be written as

t(n) = s(n) XOR t(n-4) XOR t(n-7)

The descrambler recalculates the original pattern with the same function (change

s(n) with t(n))

Period length = 2^R 1 , where R is the number of shift registers

That is, even a single 1 on the input (and all registers set to 0) will produce a 127-

bit sequence of pseudo random pattern.

This scrambler is used with 802.11b (Wireless LAN).

14


Transmission System Overview

Information

Source

Source

Coding

Channel

Coding

Line

Coding

Modulation

Information

Interpretor

Source

Decoding

Error

Detection

Descramber

Equalizer

FilterDemodulator

10110001...

Filter unnecessary bits

(Compression)

FCS and FEC (Checksum)

Bandlimited pulses

NRZ, RZ, HDB3, AMI, ...

Signal

NoiseNoise AN

AL

OG

UE

DIG

ITA

L

Coding is not coding. The above slide gives you an overview about different

coding purposes. Even modulation is sometimes called coding.

Source coding tries to eliminate redundancy within the information. Source

coders must know well about the type of information that is delivered by the

source.

Channel coding protects the non-redundant data stream by adding calculated

overhead. Typically a Frame Check Sequence (FCS) is added. Only on very

errourness and/or long-delay links a Forward Error Correction (FEC) method

might be useful. FEC requires too much overhead in most terrestial applications.

Line coding focuses on the line, that is we want the symbols to be received

correctly, even if noise and distortions are present. Furthermore line coding

provides clock synchronization as discussed earlier.

Finally modulation might be necessary in case the channel has better properties at

higher frequencies.

15


Communication Channels

Usually Low-Pass behavior Higher frequencies are more attenuated than

lower

Baseband transmission Signal without a dedicated carrier

Example: LAN technologies (Ethernet etc)

Carrierband transmission The baseband signal modulates a carrier to

match special channel properties

Medium can be shared for many users (different

carriers) e. g. WLAN

Each communication channel exhibits a low-pass behaviorat least beyond a

very high frequency. Not only is the signal attenuated; phase shifts occur and

even nonlinear effects sometimes rise with higher frequencies. The result is a

smeared signal with little energy.

In most cases the signals do not need to be modulated onto a carrier. That is, all

the channel bandwidth can be used up for this signal. We call this baseband

transmission.

Carrier and transmission put the baseband signal onto a carrier with higher

frequency. This is necessary with radio transmissions because low frequencies

have a very bad radiation characteristic. Another example is fiber optics, where

special signal frequencies are significantly more attenuated and scattered than

others.

16


Channel utilization examples

Frequency

Power

Density Baseband

Transmission

Frequency

(kHz)

Power

Density

1 2 30.3 3.4

Telephone

Channel

Frequency

Power

Density

fc1 fc2 fc3

Multiple Carriers

The above slide shows some examples for baseband and carrierband

transmission. In case we use multiple carriers we may also call it broadband-

transmission.

The third picture (bottom of slide) shows the spectral characteristic of a telephony

channel (signal). The ITU-T defined an "attenuation-hose" in great detail

(dynamics, ripples, edge frequencies, etc). As a rule of thumb we can expect low

attenuation between 300 Hz and 3400 Hz.

17


Maximal Signal-Rate

Maximal data rate proportional to channel-bandwidth B Raise time of Heavyside T=1/(2B)

So the maximum rate is R=2B, also called the Nyquist Rate

Note: We assume an ideal channel here without noise!

Bandwidth decreases with cable length As a dirty rule of thumb: BW Length @ const

But note that the reality is much more complex

Solitons are remarkable exceptions

0

1

(2B)-1

Maximum signal rate: At least

the amplitude must be reached

Since each channel is a low-pass, and some channels even damp (very) low

frequencies, data can only be transmitted within a certain channel bandwidth B.

If we put a 0 to 1 transition on the line (with ideally zero transition time), the

receiver will see a slope with a rise time of T=1/(2B).

So the maximal signal rate is T=1/(2B) in theory. In practice we need some

budget because there is noise and distortion and imperfect devices.

The longer the cable the more dramatically the low-pass behaviour. In other

words: on the same cable type we can transmit (let's say) 1,000,000,000 bits/s if

the cable is one meter in length, or only 1 bit/s if the cable is one million

kilometers in length.

It is very interesting to mention that some modern fiber optic transmission

methods violate this basic law. This methods base on so-called Soliton-

Transmission.

18


The Maximum Information Rate

What about a real channel? What's the maximum achievable information rate in presence of noise?

Answer by C. E. Shannon in 1948 Even when noise is present, information can

be transmitted without errorswithout errors when the information rate is below the channel capacitychannel capacity

Channel capacity depends only on channel bandwidth AND SNR Example: AWGN-channel

C = B log (1 + S/N)

The great information theory guru Claude E. Shannon made a great discovery in

1948. Before 1948, it was commonly assumed, that there is no way to guarantee

an error-less transmission over a noisy channel. However, Shannon showes that

transmission without errors is possible when the information rate is below the so-

called channel capacity, which depends on bandwidth and signal-to-noise ratio.

This discovery is regarded as one of the most important achievements in

communication theory.

19


Bitrate vs Baud

Information Rate: Bit/s

Symbol Rate: Baud

The goal is to send many (=as much as possible) bits per symbol

=> QAM (see next slides)

0 10 1 1 1 1 10 0 0 0 00 10 10 01 01 11

N bit/s 2N bit/s

N Baud N Baud

Baud is named after the 19th centurey French inventor Baudot, originally referred

to the speed a telegrapher could send Morse Code.

Today the symbol rate is measured in Baud whereas the information rate is

measured in bit/s.

20


Analogue Modulation Overview

t

1 0 1

Amplitude Shift Keying (ASK)

t

1 0 1

Phase Shift Keying (PSK) Frequency Shift Keying (FSK)

t

1 0 1

)2cos()( ttt tfAtg jp +=

EVERY transmission is analogue but there are different methods to put a base-band signal onto a high-frequency carrier

The most simple (and oldest) is ASK The illustrated ASK method is simple "On-Off-Keying" (OOK)

FSK and PSK are called "angle-modulation" methods (nonlinear => spectrum shape is changed!)

For digital transmission, almost always QAM is used The BER of BPSK is 3 dB better than for simple OOK

These three parameters can be modulated

The slide shows a general modulation equation. The 3 parameters of the equation describe the 3 basic modulation

types. All 3 parameters, the amplitude At, the frequency ft and the phase t, can be varied, even simultaneously. In

nature, there is no real digital transmission; the binary data stream needs to be converted into an analog signal. As

first step, the digital data will be directly transformed into a analog signal (0 or 1), which is called a baseband

signal. In order to utilize transmission media such as free space (or cables and fibers) the base signal must be mixed

with a carrier signal. This analog modulation shifts the center frequency of the baseband signal to the carrier

frequency to optimize the transmission for a given attenuation/propagation characteristic.

Amplitude Shift Keying

A binary 1 or 0 is represented through different amplitudes of a sinus oscillation. Amplitude Shift Keying (ASK)

requires less bandwidth than FSK or PSK since natura non facit saltus. However ASK is interference prone. This

modulation type also used with infrared-based WLAN.

Frequency Shift Keying

Frequency Shift Keying (FSK) is often used for wireless communication. Different logical signals are represented

by different frequencies. This method needs more bandwidth but is more robust against interferences. To avoid

phase jumps, FSK uses advanced frequency modulators (Continuous Phase Modulation, CPM).

Phase Shift Keying

The 3rd basic modulation method is the Phase Shift Keying (PSK). The digital signal is coding through phase

skipping. In the picture above you see the simplest variation of PSK, using phase jumps of 180. In practice, to

reduce BW, phase jumps must be minimized, and therefore PSK is implemented using advanced phase modulators

(e. g. Gaussian Minimum Shift Keying, etc). The receiver must use same frequency and must be perfectly

synchronized with the sender using a Phase Locked Loop (PLL) circuit. PSK is more robust as FSK against

interferences, but needs complex devices.

After understanding these modulation methods QAM shall be introduced, which is the most important modulation

scheme today for both wired and wireless transmission lines.


QAM: Idea

"Quadrature Amplitude Modulation"

Idea:

1. Separate bits in groups of words (e. g. of 6

bits in case of QAM-64)

2. Assign a dedicated pair of Amplitude and

phase to each word (A,)

3. Create the complex amplitude Aej

4. Create the signal Re{Aej ejt}

= A (cos cos t - sin sin t ) which

represents one (of the 64) QAM symbols

5. Receiver can reconstruct (A,)

22


QAM: Symbol Diagrams

Q

I

10 11

00 01

Standard

PSK

Quadrature

PSK (QPSK)

Q

I1 0

Q

I

16-QAM

Re{Ui}

Im{Ui}

1V 3V 5V

Other example:

Modem V.29

2400 Baud

Max. 9600 Bit/s

For noisy and

distorted channels

4800 bit/s

For better channels

7200 bit/s

For even better

channels

9600 bit/s

Worth to know: Simple Phase Shift Keying (PSK) which only uses two symbols,

each representing either 0 and 1. Quadrature Phase Shift Keying (QPSK) with

four symbols.

Usually the assignment of bit-words to symbols is such that the error probability

due to noise is minimized. For example the Gray-Code may be used between

adjacent symbols to minimize the number of wrong bits when an adjacent symbol

is detected by the receiver.

The above slide also shows the symbol distribution over the complex plane for

the V.29 protocol which is/was used by modems. Depending on the noise-power

of the channel, different sets of symbols are used.

14,400 bit/s requires 64 points

28,800 bit/s requires 128 points

23


Example QAM Applications

One symbol represents a bit pattern Given N symbols, each represent ld(N) bits

Modems, 1000BaseT (Gigabit Ethernet),

WiMAX, GSM,

WLAN 802.11a and 802.11g: BPSK @ 6 and 9 Mbps

QPSK @ 12 and 18 Mbps

16-QAM @ 24 and 36 Mbps

64-QAM @ 48 and 54 Mbps

It is important to understand that spread spectrum (or OFDM) techniques are always combined

with a symbol modulation scheme. Quadrature Amplitude Modulation (QAM) is a general method

where practical methods such as BPSK, QPSK, etc are derived from.

The main idea of QAM is to combine phase and amplitude shift keying. Since orthogonal

functions (sine and cosine) are used as carriers, they can be modulated separately, combined into a

single signal, and (due to the orthogonality property) de-combined by the receiver.

And since A*cos(wt + phi) = A/2{cos(wt)cos(phi) sin(wt)sin(phi)} QAM can be easily

represented in the complex domain as Real{ A*exp(i*phi)*exp(i*wt)}.

The standard PSK method only use phase jumps of 0 or 180 to describe a binary 0 or 1. In the

right picture above you see a enhanced PSK method, the Quadrature PSK (QPSK) method. While

using Quadrature PSK each condition (phase shift) represent 2 bits instead of 1. Now it is

possible to transfer the same datarate by halved bandwidth.

The QSK signal uses (relative to reference signal)

- 45 for a data value of 11




To reconstruct the original data stream the receiver need to compare the incoming signal with the

reference signal. The synchronization is very important.

Why not coding more bits per phase jump ?

Especial in the mobile communication there are to much interferences and noise to encode right.

As more bits you use per phase jump, the signal gets more closer. It is getting impossible to

reconstruct the original data stream. In the wireless communication the QPSK method has proven

as a robust and efficient technique.

24


QAM Example Symbols (1)

Note that the above QAM signals show different successive QAM-symbols for

illustration purposes. In reality each symbol is transmitted many

hundred/thousand times

25


QAM Example Symbols (2)

Note that the above QAM signals show different successive QAM-symbols for

illustration purposes. In reality each symbol is transmitted many

hundred/thousand times

These diagrams have been generated using Octave, a free Matlab clone.

26

The biggest problem

with communication

is the illusion

that it has occured.

Married?

1


Network Layers

Standardization Cruelty

This chapter introduces the layer concept widely used in data communication.

Most famous is the ISO-OSI 7-layer model, which is also discussed in great

detail here. By the way the interaction of layering and standardization is

explained.

2

The good thing

about standards is

that there are so

many to choose from

Andrew S. Tanenbaum

3


Standards

We need networking standards Ensure interoperability

Large market, lower cost (mass production)

Vendors need standards Good for marketing

Vendors create standards Bad for competitors, hard to catch up

But: Slow standardization processes freeze technology...

We need standards. Unfortunately. Otherwise, each vendor would create what

he wants and we would not be able to communicate accross networks. This

situation occured very often in history. For example the United Nations

initiated a world-wide Telephony standardization board, known as CCITT

(today ITU-T). Or in the pre-Ethernet age, many vendors built completely

incompatible LAN protocols.

Especially to force interoperability, many vendors for Internet-equipment

initiated the TCP/IP Interoperability Conference in 1987, today known as

"INTEROP".

4


Who Defines Standards?

ISO Anything

IETF Internet

ITU-T Telco Technologies

ATM Forum

Frame Relay Forum

IEEE LAN Protocols

The above slide mentions the most important standardization organizations.

The Internet Engineering Task Force (IETF) is "actually" the most important

technical organization for the Internet working groups and is organized in

several areas. Area manager and IETF chairman form the IESG (Internet

Engineering Steering Group). The IETF is also responsible to maintain the

RFCs.

5


Standards Types

De facto standards

Anyone can create them

E.g. Internet RFCs

De jure standards

Created by a standardization

organization

E.g. ISO/OSI, ITU-T

Not all standards are like the others. De facto standards are more flexible and

speed-up the implementation. Usually everybody is allowed to extend them.

The whole Internet is built on such loosely standards. Unfortunately

misinterpretations can occur. (RFCs)

De jure standards are like acts of law. For example ITU-T standards explain

nearly every detail implementers may ask.

6


Note

Standardization is applied

to network layersnetwork layers

and interfaces interfaces

between them

The above sentence leads us to network layers. Break big problems into smaller

ones and write standards for them ("divide and conquer"). Of course the

interfaces between the layers must be standardized too. Eventually, multiple

developers can work on different parts of the whole story.

7


Network Layers

Divide task of communication in multiple sub-tasks

Hierarchically organized Each layer receives services from the

layer below

Each layer serves for the layer above

Good for interoperability Capsulated Entities and Interfaces

But increases complexity

Network layers are an abstraction to hide complexity. Layers are organized

hierarchically, that is there is a predefined command direction. Imagine what

would happen if we have a democratic model?

Note that network layers force a more complex development. Many high-

performant communication technologies have been developed in an ad-hoc act,

or alternatively consists of only a few layers.

8


Where to Define Layers

Group functions (services) together

When changes in technology occur

To expose services

To allow changes in protocol and HW

To utilize existing protocols and HW

A good layering structure requires a intelligent grouping of functions. Ideally,

technology improvements can be implemented immediately.

For example the X.25 packetizing algorithm, which is written in software and

part of a network driver of the operating system can remain untouched, while

the serial line hardware can be updated, and vice versa.

9


The ISO/OSI Model

International Standards Organization (ISO)

International agency for the development of

standards in many areas

Founded 1946

Currently 89 member countries

More than 5000 standards until today

1988 US Government OSI Profile (GOSIP) Requires Government products to support OSI

layering

The ISO standardized anythingcharacter sets, paper sizes, screws, ..., and

network layers. In 1988 the US Government required any communication

device to comply with the ISO/OSI model (GOSIP). Note that the non-OSI

Internet was built much earlier, so many people expected the end of the

Internet. But the Internet (which was created as nuclear-bomb resistant) not

only survived the ISO/OSI model but also displaced many OSI-compliant

protocols, such as CLNP.

Similarly, in Europe the "European Procedurement Handbook for Open

Systems" (EPHOS) had been released.

10


Purpose

OSI model describes communication

services and protocols

No assumption about

Operating system

Programming Language

Practically, the OSI model

Organizes knowledge

Provides a common discussion base

Although every book of data communication mentions the ISO/OSI 7-layer

model it is not that important in the real world: most technologies do not

comply to this model. It is merely a reference model so that we can refer to it

when we want to explain certain functions in our protocols. From this point of

view the OSI model is indeed important today.

11


OSI Basics

Point-to-Point, no shared media

Nodes are called End Systems (ES)

Intermediate Systems (IS)

Each layer of the OSI model detects and handles errors (FCS)

Dumb hosts and intelligent network Compared with Internet: dumb network,

intelligent hosts

The original OSI modes was created for point-to-point connections only (for

example there was no specification for LAN-like shared media originally).

12


The OSI Truth

OSI model was created before

protocols

Good: Not biased, general approach

Bad: Designers had little experience, no

ideas in which layers to put which

functionality...

Not widespread (complex,

expensive)

But serves as good teaching aid !!!

Although the OSI Model was created before the protocols, and so the complete

model is very complex and not practically elaborated, its widely used today to

define and category most of the important protocols. OSI is not biased because

this reference framework is not associated with any particular vendor

philosophy. OSI represents a general approach for describing data

communication procedures but this property is often considered as a big

disadvantage, because practical implementations typically can be described

with a much simpler model and on the other hand the OSI architects had only

little experience with real life implementations. Therefore, genuine OSI

protocols are not really widespread today, because of its complexity.

Nevertheless, the OSI model serves as reference frame when discussing or

learning about protocols.

13


The 7 OSI LayersThe 7 OSI Layers

Application Layer

Transport Layer

Network Layer

Data Link Layer

Physical Layer

Session Layer

Presentation Layer

Application Layer

Transport Layer

Network Layer

Data Link Layer

Physical Layer

Session Layer

Presentation Layer

System A System B

Sender Process Receiver Process

Because the communication between different systems can be a very complex

task, OSI splits the communication aspects into smaller tasks. All layering is

based on the OSI reference Model, which defines tasks and interactions of

seven layers.

The users data moves from the first layer (Application Layer) through all other

layers. When two systems communicate with each other, then only the different

layers talk. The application layer only talk with the application layer or the

network layer only communicate with the network layer of system B. We can

talk about a parallel communication between the layers. Every layer works for

its own, it is not interested what the other layer does.

14


Physical Layer

Mechanical and electrical

specifications

Access to physical medium

Generates Bit stream

Line coding and clocking

Examples LAN: Ethernet-PHY, 802.3-PHY

WAN: X.21, I.400 (ISDN),

RS-232

Application Layer

Transport Layer

Network Layer

Data Link Layer

Physical Layer

Session Layer

Presentation Layer

The Physical Layer generates the bit stream. This layer provides access to the

physical medium by applying line coding (NRZ, Manchester, etc),

synchronization (clocking, PLL), but also includes mechanical specifications.

Layer 1 also can activate or deactivate the links between end systems (link

management).

The physical part of the Ethernet NIC is called "PHY" and is perhaps the most

complex entity of Ethernet because the PHY consists of a number of sublayers

that care for interoperability with different Ethernet speeds (10, 100, 1000,

10000 Mbit/s) and codings (Manchester, 4B5B, 8B10B, ...). Note that there is

a fundamental difference between "Ethernet" and IEEE "802.3": these are two

separate LAN specifications but typically implemented on the same NICthey

just share the same topology and use the same media access strategymost

people are not aware of that.

The X.21 is a typical and widely available interface on a Cisco router. The

ISDN-layer 1 is specified in the ITU-T standard I.400 and describes both a 192

kbit/s synchronous multiplexing interface capable to transport 2 B channels and

one D channel and secondly a high speed 2.048 Mbit/s interface capable to

carry 30 B channels and one D channel. These ISDN specifics are presented in

the N-ISDN chapter in more details. The old well-known Recommended

Standard (RS) 232 specifies the classical serial interface found on many PCs

and other peripheral devices.

15


Link Layer

Reliable transmission of frames between two NICs

Framing

FCS

Physical Addressing of NICs

Optional error recovery

Optional flow control

Examples: LAN: 802.2

PPP, LAPD, LAPB, HDLC

Application Layer

Transport Layer

Network Layer

Data Link Layer

Physical Layer

Session Layer

Presentation Layer

The data link layer builds the frame. In that way, framing or frame

synchronization is the most important thing on layer 2. Where is the beginning

of the frame ? Where is the end ? With a special Bit-Code the layer 2 protocols,

such as HDLC or PPP, guarantee the framing of the data. Thats important for

the MTU (maximum transfer unit).

Also frame checking, correction of transmission errors on a physical link, is

implemented on layer 2. There are also a physical address of the network

interface cards. This address is transported with the data link layer too (e.g.

MAC-Address with Ethernet).

Error recovery and flow control may be realized in connection-oriented mode.

16


Network Layer

Transports packets between networks

Provides structured addresses to name networks

Fragmentation and reassembling

Examples: CLNP

IP, IPX

Q.931, X.25

Application Layer

Transport Layer

Network Layer

Data Link Layer

Physical Layer

Session Layer

Presentation Layer

The network layer builds the so-called "packet". Layer 3 transports the packets

between the different networks. Therefore layer 3 needs structured and routable

addresses to find the right networks. IP is the most important Layer 3 protocol

today (IPv4 has a structured 4 byte address). The OSI Connectionless Network

Protocol (CLNP) is another example for a layer-3 protocol but it is not so

widely used today, except some Telcos and Carriers use it for internal

purposes. IPX has been developed by Novell in order to extend Novell

networks over different data-link layer worlds. Q.931 is the ISDN layer 3

carried over the D-channel and is used for signaling purposes. Basically Q.931

conveys the telephone numbers and other service parameters. The classical

packet-switched WAN standard X.25 actually specifies only the layer 3 of this

technology and is used to set up a number of virtual calls over an asynchronous

link layer (LAPB).

17


Transport Layer

Reliable transport of segments between applications

Application multiplexing through T-SAPs

Sequence numbers and Flow control

Optional QoS Capabilities

Examples: TCP (UDP)

ISO 8073 Transport Protocol

Application Layer

Transport Layer

Network Layer

Data Link Layer

Physical Layer

Session Layer

Presentation Layer

The transport layer is necessary to build a logical connection to the application

in order to send data in so-called "segments". With the help of port numbers

(by TCP and UDP), a layer 4 protocol guarantees the transport of the segments

to the right application. These port numbers are called T-SAPs in the OSI

world. The transport layer optionally takes care about flow control, reliable

transmission between end systems, and is most important for QoS capabilities.

Flow control requires connection oriented mode.

18


Session Layer

Provides a user-oriented

connection service

Synchronization Points

Little capabilities, usually

not implemented or part of

application layer

Telnet: GA and SYNCH

FTP: re-get allows to continue

an interrupted download

ISO 8327 Session Protocol

Application Layer

Transport Layer

Network Layer

Data Link Layer

Physical Layer

Session Layer

Presentation Layer

The Session Layer coordinates and controls dialogue between different end

systems. This layer is only seldom or sparsely implemented. For example a

Telnet server gives the sending permission to the Telnet client via a Go Ahead

(GA) sequence. Using the BRK-Key, a SYNCH sequence is triggered and the

server must synchronize with the client by flushing the buffered stream. FTP

keeps track of the data blocks transmitted and is able to continue an interrupted

session from this checkpoint on.

Session protocols are important with telephony applications such as H.323

which employs H.225 to establish sessions. Another example is the IETF

Session Initiation Protocol (SIP). The ISO 8327 is an OSI basic connection

oriented session protocol specification.

19


Presentation Layer

Specifies the data

representation format for the

application

Examples:

MIME (part of L7) and

UUENCODING (part of L7)

ISO: ASN.1 and BER

Application Layer

Transport Layer

Network Layer

Data Link Layer

Physical Layer

Session Layer

Presentation Layer

The layer 6 is responsible for common language between end systems. The

presentation layer specifies the "meaning" of the data and how each byte

should be interpreted.

In the Internet the presentation layer uses ASCII coding and the meaning of the

data is specified by a so-called "Multipurpose Internet Mail Extension"

(MIME) header. MIME is used by SMTP (Email) and HTTP (Web browsing)

for example. UUENCODING is one example of how to transform 8-bit-bytes

into 7-bit-bytes and it is typically used with Internet Mail attachments. The

ISO/OSI world generally uses the "Abstract Syntax Notation Language

Number One" (ASN.1) as common presentation layer. This language is used to

specify data structures and contents. On the wire the data is transmitted using

the "Basic Encoding Rules" (BER).

20


Application Layer

Provides network-access for

applications

Examples: ISO 8571 FTAM File Transfer

Access + Management,

X.400 Electronic Mail, CMIP

SMTP, FTP, SNMP, HTTP,

Telnet, DNS,

Application Layer

Transport Layer

Network Layer

Data Link Layer

Physical Layer

Session Layer

Presentation Layer

The Application layer supports user with common network applications. For

example: file transfer or virtual terminals. Layer 7 also supports basic network

procedures in order to implement distributed applications (e.g. transaction

systems). Note that the application layer is not identical with the application!

The application itself "sits" upon the application layer and uses the service

primitives provided by the application layer to access the network.

Application layer protocols either use "inline" or "inband" control sequences

(as it is used with Telnet), where control bytes are mixed with the data stream,

or it might use a predefined frame structure, consisting of header and body.

Another method is to open a dedicated logical control connection only to

exchange control information (as it is used with FTP).

21


Encapsulation Principle

L7

L4

L3

L2

L1

L5

L6

DATA

DATA

A-PDU

P-PDU

S-PDU

T-PDU

N-PDU

7

4

3

2

1

5

6

DATA

101000111010010110100101001010000100101010001010101010101010010110001001010101010100101111100000101010

L-PDU or "Frame"

N-PDU or "Packet"

T-PDU or "Segment"

S-PDU

P-PDU

A-PDU

The data moves through all 7 layers. Every layer add his own header. The data

with layer 4,5,6 and 7 header is called segment. A segment plus layer 3

header is called packet. The so called frame (data plus six headers) will be

transport over layer 1 to the destination system. Then the frame will moves

through all 7 layers again, and in each station a header will be removed.

22


Practical Encapsulation

Ethernet Frame

IP Packet

TCPSegment

HTTPMessage

HTMLWebpage

The idea of encapsulation is fundamental in the data communication world.

Adjacent layers encapsulate or decapsulate information by adding/removing

additional "overheads" or "headers" in order to implement layer-specific

functionalities. The whole process can be regarded as Matroschka-puppet

principle.

In our example let's suppose a webserver sends a webpage (HTML code) to a

client. The webpage is carried via the Hyper Text Transfer Protocol (HTTP)

which provides for error and status messages, encoding styles and other things.

The HTTP header and body is carried via TCP segments, which are sent via IP

packets. On some links in-between, the IP packets might be carried inside

Ethernet frames.

23


Internet Encapsulation

HTTPHeader

HTTP-Data

HTML-Content

(Webpage)

TCPHeader

TCP-Data

IPHeader

IP-Data

Will reach the next

Ethernet DTE Eth

HeaderEthernet-DataEth

Trailer

Will reach the target host

Will reach the target application

This is what the application wants

This is what the user wants

In our example let's suppose a webserver sends a webpage (HTML code) to a

client. The webpage is carried via the Hyper Text Transfer Protocol (HTTP)

which provides for error and status messages, encoding styles and other things.

The HTTP header and body is carried via TCP segments, which are sent via IP

packets. On some links in-between, the IP packets might be carried inside

Ethernet frames.

24


OSI Speak (1)

Entities Anything capable of sending or

receiving information

System Physically distinct object which

contains one or more entities

Protocol Set of rules governing the exchange of

data between two entities

Entities:

Any hardware or software module that acts upon a single layer is called an

"entity". Several entities exist peer to peer within a given layer and are capable

to communicate which each other. This type of communication is referred as

"horizontal" communication -- this is actually what we mean when we talk

about a "protocol".

System:

Several entities make up a "system". For example a PC is a "system" because

it consists of the entities Ethernet PHY entity, MAC entity, LLC entitiy, IP

entity, TCP entity, and several L7 entities. A system is merely a term that

reflects the physically separation of groups of entities.

Protocol:

We already described the meaning of protocol above together with the

definition of a entity, but a protocol can be explained more simple: A protocol

is a set of rules that are necessary to exchange data in an ordered and

unmistakable way.

25


OSI Speak (2)

Layer

A set of entities

Interface

Boundary between two layers

Service Access Point (SAP)

Virtual port where services are passed

through

Layer:

A "layer" in the OSI jargon is a set of entities--but do not confuse layers with systems! The

entities of a layer reside on the same hierarchy level and a single layer comprises several

systems. On the other hand a system comprises several layers but typically only one (or a

limited number) of entities are available on each layer of a system. For example: In order to

communicate in the Internet, all devices must support layer 3 (the IP layer). That is, each

system must provide at least one IP-entity.

Interface:

An "interface" is simply the logical boundary between two layers. Note that interfaces are

typically not physically visible because they represent the boundary between two layers at a

whole. The local representation of an interface is called a "Service Access Point" or SAP. The

Service Access Point is one of the most frequently used terms in data communication and

simply reflects the piece of hardware or software that acts as an interface between two layers.

The previously OSI-interface is meant globally, while the SAP has local meaning, i. e. at one

system. A SAP is a practical term, in some technologies such as IEEE 802.2 it is just a field in

the header indicating the destination and source layer. If you use an Ethernet NIC with an AUI

interface, than this electrical interface can be also considered a SAP because "service

primitives" are passed through this interface. Service Primitives are explained below...

Service Access Point:

An "Interface Data Unit" (IDU) is practically spoken the piece of data that is passed through a

SAP to the next layer's entity. It contains ICI and SDU which is described below. When an

IDU is passed through a SAP to the next layer, this layer extracts and processes the Interface

Control Information (ICI).

26


OSI Speak (3)

Interface Data Unit (IDU)

Data unit for vertical communication

(between adjacent layers of same

system)

Protocol Data Unit (PDU)

Data unit for horizontal communication

(between same layers of peering

systems)

Interface Data Unit:

An "Interface Data Unit" (IDU) is practically spoken the piece of data that is

passed through a SAP to the next layer's entity. It contains ICI and SDU which

is described below. When an IDU is passed through a SAP to the next layer,

this layer extracts and processes the Interface Control Information (ICI).

Note that data is passed through a SAP using "service primitives". Service

primitives are functions that are implementation specific (for example an API)

and are used to pass data from one layer to another on the same system. These

service primitives actually pass on these IDUs.

Protocol Data Unit:

The SDU actually represents the payload plus headers for upper layers. The

SDU is transported horizontally with an header used at this layer. Both SDU

and Header is called a "Protocol Data Unit" (PDU). The PDU is the most

often used term of all these terms mentioned here. At least you should

remember the PDU.

27


OSI Speak (4)

Interface Control Information (ICI)

Part of IDU

Destined for entity in target-layer

Service Data Unit (SDU)

Part of IDU

Destined for further communication

Contains actual data ;-)

28


OSI Speak Summary (1)

(N) Layer

(N+1) Layer

(N-1) Layer

Interface

Interface

(N) Layer

Entity

(N+1) Layer

Entity

(N+1) Layer

Entity

(N-1) Layer

Entity

(N-1) Layer

Entity

"Protocol"

Service Access Point (SAP)

Service Primitives

Service Primitives

(N) Layer

Entity

The ISO/OSI model defines four service primitives: request, indication,

response and confirm.

Note that the service primitives are only used for vertical communication.

29


OSI Speak Summary (2)

(N) Layer

(N+1) Layer

Interface

(N) Layer

Entity

(N+1) Layer

Entity

(N) Layer

Entity

ICI SDUIDU

ICI SDU

SDU NH

N-PDU

SAPVertical

Communication

Horizontal

Communication

30


Layer 1 Devices

Adapts to different physical

interfaces

Amplifies and/or refreshes the

physical signal

No intelligence

Repeater, Hub,

NT1

Application

Transport

Network

Data Link

Physical

Session

Presentation

Application

Transport

Network

Data Link

Physical

Session

Presentation

Repeater

To connect different system with each other we need special devices. If you

want to connect two systems only per physical layer you need a so called hub

or repeater.

This kind of devices are not intelligence and only used to amplifies or refresh

the physical signal, or to connect systems with different physical interfaces.

31


Layer 2 Devices

Filter/Forwards frames according Link Layer Address

Incorporates Layer 1-2

LAN-Bridge ("Switch")

Application

Transport

Network

Data Link

Physical

Session

Presentation

Application

Transport

Network

Data Link

Physical

Session

Presentation

Bridge

A so called bridge or switch is a device to connect systems per data link

layer. This kind of devices determine the physical layer and can forward

datagram's according the link layer address. For example: MAC address with

Ethernet. Note that a bridge utilizes two or more physical layer entities (NICs)

that is a bridge is able to convert encodings and signal-rates.

Note the difference between bridge and switch: A bridge is implemented in

software, whereas a switch is built in hardware. Today only switches are used,

because they are much faster.

32


Layer 3 Devices

"Packet Switch" or "Intermediate System"

Forwards packets to other networksnetworks according structuredstructured address

Terminates Links

Router,WAN-Switch

Application

Transport

Network

Data Link

Physical

Session

Presentation

Application

Transport

Network

Data Link

Physical

Session

Presentation

Router

The most important device in the Internet is the so called router. A router

consists of several layer 1 and layer 2 entities and a single layer 3 entity, thus it

can forward packets to other networks according structured addresses

(remember IP-Addresses). By terminating layer 1 and 2 a router is able to

connect total different network technologies with each other. For example: on

one side there is Ethernet on the other side ATM.

33


A Practical Example

Physical(Twisted Pair)

Physical(Serial Line)

Physical(Fiber Ring)

Link(Ethernet)

Link(HDLC)

Link(FDDI)

Network(IP)

Transport(TCP)

Netscape

Browser

Apache

Webserver

MAC

Address

MAC

Address

Simple or

dummy Address

IP A

ddre

ssIP

AddressP

ort

Num

ber

Port N

um

ber

What is my

destination application?

Where is my

destination network?

Just move this frame

to the next NIC

In the picture above you see a good example in which symbolic way the

different layers talk with each other. The link layer only searches for the right

NIC address. IP only wants to the destination network, and TCP is the protocol

to communicate between applications. Most importantly, notice that packets

are sent over different link layer technologies such as Ethernet, HDLC, or

FDDI. Exactly this is the reason why a common network layer is needed to

allow communication over different "networks" (=links).

Don't be confused about the different usages of the term "network". People say

"network" and mean "bunch of devices interconnected with each other". To be

more exact, a network is identified by a unique network identifier, such as the

network-ID of the IP-address. Since a contiguous link layer implementation

(such as an Ethernet LAN) can have assigned a single IP net-ID, each link can

be regarded as network.

34


Padlipsky's Rule

If you know what

you're doing, three

layers is enough. If

you don't, even

seventeen won't help.

Until now we discussed the famous OSI seven layer reference model, but real

implementations typically consist of a subset of this 7-layer model. On the one

hand, not all OSI layers are necessary in real-world applications, on the other

hand, many important technologies had been created before the OSI standard.

35


Stevens 4-Layer Model

Transport Layer

Network Layer

Data Link Layer

Process Layer

Transport Layer

Network Layer

Data Link Layer

Process Layer

Equivalent to the DoD Model (Internet)

The picture above shows the W. Stevens 4 layer model which is used also in

the Internet. The Internet layer model is also called "Department of Defense"

(DoD) model.

36


Tanenbaum 5-Layer Model

Application Layer

Transport Layer

Network Layer

Data Link Layer

Physical Layer

Application Layer

Transport Layer

Network Layer

Data Link Layer

Physical Layer

The famous computer scientist Andrew S. Tanenbaum defined a more practical

approach utilizing five layers. Other than the DoD or Stevens 4-layer model

the physical specifications are defined in a separate layer.

37


Summary

Network layers ensures

interoperability and eases

standardization

ISO/OSI 7 layer model is an

important reference model

Practical technologies employ a

different layer set, but it's always

possible to refer to OSI

The Internet perspective is implement it,

make it work well, then write it down.

The OSI perspective is to agree on it,

write it down, circulate it a lot and now

we'll see if anyone can implement it after

it's an international standard and every

vendor in the world is committed to it.

One of those processes is backwards,

and I don't think it takes a Lucasian

professor of physics at Oxford to figure

out which.

Marshall Rose, "The Pied Piper of OSI"

39


Quiz

Explain layer-2 capabilities!

What could be the task of a layer-4

device ?

What is a "gateway"?

How does the (N) layer tell (N+1)

layer that it has data to hand over ?

Why have OSI protocols not been

successful on market ?


Hints

Q1: Framing, Protection, Access,...

Q2: Layer 4 device might deal with QoS, sequencing and flow control

Q3: According to OSI a layer 1-7 device, according to IETF a router.

Q4: Using Service Primitives (Indicate)

Q5: OSI is too complex and general, several fields in headers might have variable length, sometimes ignores byte- and word-delineation, ...

1


Protocol Principles

Framing, FCS and ARQ

2


Link Layer Tasks

Framing

Frame Protection

Optional Addressing

Optional Error Recovery

Connection-oriented or

connectionless mode

Optional Flow Control

The Data-link layer of the OSI model is the first layer above the physical layer

that is used to perform logical tasks like:

Framing is the task of packing the information of higher layers to provide for

example start and end of packet detection plus some optional features

which will be discussed on the following slides

Frame protection is used to detect possible errors during data transmission

Addressing is optional and is normally only used in point-to-multipoint Data-

link technologies.

Error recovery can be used to allow packet retransmissions if data errors are

detected by the frame protection mechanism

The Data-link can be driven in connection oriented mode or connection-less

mode. Newer technologies mainly use the connection-less mode because the

connection-oriented functionality is typically provided by higher OSI layers

e.g. TCP on OSI layer 4.

Flow control can be used to prevent buffer overflow situations on the receiver

side

3


Building a Frame

DATAControl FCS EDSDPreamble

Consists of

Data

Metadata (Header or "Overhead")

Requires synchronous physical

transmission (PLL)

Arbitrary frame lengths

A basic layer 2 transmission frame consists of following components:

Preamble - is used to provide synchronization between the sender and the

receiver transmission clock. This is necesarry to allow the detection of the

single bit borders.

SD - Start Delimiter is needed to detect the actual beginning of the layer 2

transmission frame. From this point on data is fed from the physical layer into

the receive buffer.

Control Field - provides optional addressing, connection establishment, error

recovery and flow control

Data - is the payload provided by higher OSI layers

FCS - Frame Check Sequence is used for error detection

ED - End Delimiter is used to determine the end of the layer 2 frame

4


Preamble


Enables PLL synchronization Typically a 0101010...-pattern

Example: 8 Byte preamble in Ethernet frames

Note: Only necessary when sender- and

receiver-clock are not synchronized

between frames

Asynchronous physical layer

The purpose of the Preamble is to lock the receiver clock towards the sender

clock by the help of Phase Locked Loop (PLL) circuits. The Preamble is

different depending on the type of Data-link technology that is used.

In Ethernet technology for example the bit pattern consists of 62 clock changes

between a logical 1 and a logical 0, followed by two logical 1s to indicate the

start of frame.

The Preamble is obviously only needed for synchronous physical layers. In the

case that an asynchronous physical layer is used, e.g. COM port on PC or async

serial interface on a router, the Preamble is not needed. Because sender and

receiver clocks don t need to be synchronized.

5


Frame Synchronization


Beginning and ending of a frame is

indicated by SD and ED symbols

bit-patterns or code-violations

lenght-field can replace ED (802.3)

Idle-line can replace ED (Ethernet)

Also called "Framing"

Starting

Delimiter

Ending

Delimiter

There are different methods available to indicate the start and the end of a data

frames. The simplest method is the use of a special bit pattern. In the HDLC

protocol and its derivates the bit pattern 01111110 is used to indicate start

and end of frame.

In Token Ring technology code violation is used. Code violation is an intended

brake in the rules of coding.

In Ethernet technology the SD is indicated by the bit pattern 11 following the

Preamble. But the end of the frame is indicated by an idle line, this means

silence on the wire for a specified period of time.

Optionally the ED can be omitted if an length field is present inside the Data-

link frame. In this case the end of frame can be calculated by counting the

number of bytes received.

6


Protocol Transparence

What, if delimiter symbols occur

within frame ?

Solution:

Byte-Stuffing

Bit-Stuffing

DATAControl FCS EDSDPreamble ED SD

! !

In the case that a special bit pattern is used to indicate the start and end of a

frame, it is necessary to prevent this pattern inside the data portion of the

frame. Otherwise this would lead to frame misinterpretation.

There are two principle methods to achieve this goal either by modifying single

bits of the data stream (bit-stuffing) or by replacing the whole byte (byte-

stuffing).

7


Byte Stuffing

Some character-oriented protocols divide data stream into frames Old technique, not so important today

Data Link Escape (DLE) character indicates special meaning of next character

A B C DLE DLE E F G ETX H I STX HSTX ETXDLE DLE

A B C DLE E F G ETX H I STX H

Data to send:

Typically, character-oriented protocols use asynchronous transmission (start-

stop bits), so the receiver gets a bunch of characters but has to find the frames

in it. In this case special control characters Start of Text (STX, 0x02) and End

of Text (ETX, 0x03) are used to indicate start/end of a transmission frame.

Obviously STX and ETX must not appear inside the date portion, so an

additional special control character Data Link Escape (DLE, 0x10) is used.

STX and ETX are only interpreted as control characters if the DLE pattern is

found in front of them. This means if STX or ETX bit pattern are found inside

the data stream, with the DLE pattern in front of them, they are interpreted as

data.

If the DLE pattern itself would appear inside the data portion it is doubled to

indicate that it is only data.

8


Bit Stuffing

Used in bit-oriented protocols Used by most protocols Bits represent smallest transmission unit

HDLC-like framing: 01111110-pattern Rule:

Trasceiver-HW inserts a zero after five ones Receiver rejects each zero after five ones

010011111000111111100101100110

Data to send:

0100111110000111110110010110011001111110 01111110

In HDLC technology and its derivates bit stuffing is used to prevent the

appearing of the SD, ED pattern inside the data frame.

This means there is a common rule between sender and receiver that after every

fifths 1 an additional logical 0 will be inserted by the sender in the data

stream. The receiver itself removes all logical 0s following five logical 1s.

9


Code Violations

Manchester

AMI

DifferentialManchester

Code violation is an intended hurt of the rules of coding. It can be used for

signaling purposes.

In Token Ring systems for example the differential manchester code is used.

Violations of the differential manchester code are used for the SD and ED

patterns to indicate the start and the end of frame.

The differential manchester code violation symbols are called J and K. The

code violation in the differential manchester code is achieved by omitting the

change from 1 to 0 or from 0 to 1 in the middle of a pulse.

10


Frame Protection

A frame check sequence (FCS) protects the integrity of our frame From Sunspots, Mobile-Phones, Noise,

Heisenberg and others

FCS is calculated upon data bits Different methods based on mathematical

efforts: Parity, Checksum, CRC

Receiver compares its own calculation with FCS

DATAControl EDSDPreamble FCS

Protected

The Frame Check Sequence (FCS) allows the receiver to detect possible errors

in the data stream.

The FCS is calculated by the sender and is attached at the end of the frame. The

calculation of the FCS is typically performed in hardware.

There are many different technologies available to calculate an FCS like:

Parity bit calculation

XOR operation

Modulo operations

Cycle Redundancy Check (CRC)

Forward Error correction (FEC)

11


FCS Methods

Parity Even (100111011) or odd (100111010)

parity bits

Examples: Asynchronous character-transmission and memory protection

Checksum Sum without carries (XOR operation)

Many variations and improvements

Examples: TCP and IP Checksums

The simplest method of FCS technologies is the parity bit calculation. This

method is typically used in asynchronous character based transmission systems.

One parity bit is computed for each single character. The first two least

significant bits are XORed together and the output of this operation is then

XORed with the next more significant bit, and so on. The output of the final

operation represents the required parity bit, which can be even (1) or odd (0).

Obviously the parity check can only protect against single bit errors. A two bit

failure for example in the opposite direction 1 to 0 and 0 to 1 cannot be

detected by the parity check.

For packet based transmission systems a similar method to calculate a

checksum can be used. Instead of a single bit, 16 or 32 bit long words are used

in combination with the XOR operation.

The XOR operation is also known as a modulo-2 adder, since the output of the

XOR operation between two binary digits is the same as the addition of the two

digits without a carry bit.

12


Checksum Example: ISBN

100% Protection against Single incorrect digits

Permutation of two digits

Method: 10 digits, 9 data + 1 checksum

Each digit weighted with factors 1-9

Checksum = Sum modulo 11

If checksum=10 then use "X"

ISBN 0-13-086388-2

0*1+1*2+3*3+0*4+8*5+6*6+3*7+8*8+8*9 = 244

244 modulo 11 = 2

ISBN stands for International Standard for Book Numbering.

The ISBN code which you may use to order your favourite books is protected

by an checksum as well. This checksum is built on the basis of an modulo 11

operation.

The ISBN code itself consists of 10 digits where the first nine digits represent

the book code and the tens digit is the checksum used for error detection.

The ISBN checksum allows the detection of a single bit failure or the

permutation of two bits with a probability of 100%.

13


Cyclic Redundancy Check

CRC is one of the strongest methods

Bases on polynomial-codes

Several standardized generator-

polynomials

CRC-16: x16+x15+x2+1

CRC-CCITT: x16+x12+x5+1

Checksums which are built on basis of the XOR operation do not provide a

reliable detection scheme against error bursts.

Therefore the Cycle Redundancy Check is based on mathematical polynomial

equations and it is capable to detect even bursts of erroneous bits inside the

data stream.

A single set of check bits is generated using an polynomial equation for each

transmitted frame and appended to the tail of the frame by the sender. The

receiver then performs the same computation on the complete frame and

compares the received checksum with its own calculated checksum.

There are a lot of different standardized polynomial equations like the CRC-16,

CRC-32 or the CRC-CCITT equations.

The CRC-16 check for example will detect all error bursts of less than 16 bits

and most error bursts greater than 16 bits. The CRC-16 and the CRC-CCITT

are mainly used in WAN technologies, while the CRC-16 is mainly used in

LAN environments.

It is quite simple to implement these CRC checks in Hardware with the use of

16 or 32 bit long shift registers.

14


Forward Error Correction

Required for "extreme" conditions High BER, EMR

Long delays, space-links

Introduces much redundancy Example: Reed-Solomon codes,

Hamming-codes

1 1 1 0

0 1 0 0

0 0 1 1

0 1 0 1

1

1

0

0

1 1 0 0 0

1 1 0 0

0 1 0 0

0 0 1 1

0 1 0 1

1

1

0

0

1 1 0 0 0

All till now discussed error protection technologies lead to an drop or optional

retransmission of the corrupted data fame. With the help of Forward Error

Correction (FEC) technologies it is possible to fix the faults inside a data frame

up to a certain extend.

One error correction coding scheme is the Hamming FEC scheme. In this

scheme the data bits plus the additional check bits are called the codeword. The

minimum number of bit positions in which two valid codewords differ is

known as the Hamming distance of the code.

It can be shown that to correct n errors we need a code with a Humming

distance of 2n + 1.

In practice the number of check bits needed for error correction is much larger

than the number of bits needed just for error detection. Therefore in most

transport systems ARQ techniques are still used for error correction.

FEC technologies are mainly used in nische technologies, e.g. communication

with space probes, where the RTT is very high and sometimes only a

unidirectional communication channels exists.

15


Control Field

Contains protocol information

Addressing

Sequence numbers

Acknowledgement Flag

Frame Type

SAP or Payload Type

Signalling information

The contents of the control field depends on the tasks that need to be

performed by the Data-link protocol.

So the control field could contain:

Address information for addressing especially in point to multipoint

environments

Sequence numbers that can be seen like serial numbers for each single frame

Acknowledgement Flags to indicate that the data was received properly

Frame Type information to indicate whether its a frame that carries data or

control information

Service Access Point (SAP) or payload type information to indicate what is

transported by the frame

Signaling information in case of connection oriented protocols to build up an

connection

16


Connection-Oriented Protocols

Different definitions

Some say "protocols without

addressing information" and think of

circuit-switched technologies

Some say "protocols that do error

recovery"

Correct: "protocols that require a connection establishment before sending data and a disconnection procedure when finished"

In the past an connection oriented protocol was very often seen as a protocol

that performs a connection setup procedure and supports error recovery and

flow control.

But newer technologies like frame relay typically do not carry address

information, do not need an connection setup procedure in case of PVC service

and do not support error recovery and flow control. But they are still

connection oriented.

So we may say there are two different types of connections temporary (like

SVC) and permanent (like PVC).

For temporary connections we obviously will need addressing and connection

setup procedures.

For permanent connections a virtual circuit identifier is enough.

In both cases temporary or permanent connections we may optionally support

error recovery and flow control.

17


CO-Protocol Procedures (1)

time time

Connection Request

Connection Established

DATA

Disconnection Request

Disconnected

Station A Station B

In connection oriented protocols a connection is established before data is

allowed to flow.

The connection establishment is done by special control frames like connection

request and connection established.

Then we find the data exchange phase which may use error recovery and flow

control. When the data transmission is finished we have special control frames

like disconnect request and disconnected to tear down the connection again.

18


CO-Protocol Procedures (2)

time time

Keepalive

Keepalive ACK

Other synonyms:

"Hello" or

"Receiver Ready"

.

.

.

.

.

.

.

.

(Connection

already

established)

Connection oriented protocols use special keep-alive procedures to make sure

the connection is up in periods where no data frames are transmitted.

19


CO-Issues

Establishment delay

Traffic desriptor during

establishment (QoS)

Additional frame types necessary

Connection establishment

Disconnect

Keepalive

ARQ possible (Error Recovery)

Obviously it takes some time to establish a connection before the data is

allowed to flow. But the establishment delay is typically in the range of

milliseconds. Even ISDN provides a connection establishment worldwide in

less than one second.

A traffic descriptor may be used optionally for all technologies that support

Quality of Service features. The traffic descriptor holds the information about

the service parameters, e.g. delay, burst size etc, that should be used for this

connection.

There are also separate frame types defined for connection establishment,

disconnect procedures and keep-alives.

The use of Automatic Repeat Requests (ARQ) is optional depending on the

technology used. The purpose of the ARQ is to provide error recovery in case

of transmission failures.

20


Automatic Repeat Request

ARQ protocols guarantee correct

delivery of data

Receiver sends acknowledgements

Acknowledgements refer to sequence

numbers

Missing data is repeated

When do we need this?

For most data traffic (FTP, HTTP, ...)

Not for real-time traffic (VoIP)

ARQ techniques are used to allow data retransmissions in the case of

transmission errors and packet loss.

With the help of sequence numbers (serial number of a data packet), applied by

the sender, the receiver is able to detect packet loss and is further able to

acknowledge properly received frames.

Reliable transmission techniques are mainly used for data traffic e.g. SMTP,

HTTP, FTP etc, because we don t want to receive corrupted mails or html

pages.

For real time traffic like Voice over IP systems we prefer unreliable

transmission techniques, because it makes no sense to retransmit a lost word a

few milliseconds later again. This would destroy the harmony in the speech

even more than the lost word. For real time systems Forward Error Correction

systems would be much more useful.

21


ARQ Variants

Idle-RQ Selective ACK

Positive ACK

GoBackN

SREJ

Continuous-RQ

There are two major families of ARQ requests the idle-RQ system and the

continuous-RQ system.

The continuous-RQ system consists of four different flavors. The Selective

ACK, Positive ACK, GoBackN, the Selective Reject (SREJ) method and

sometimes even a mix out of these basic methods.

22


Idle-RQ

Sender Receiver

Data

Ack

Data

Ack

Ack

Ack

Data

Data

The idle-RQ technique is a stop and go protocol, this means when the sender

has sent out one data frame he must wait for the according acknowledgement

before he is allowed to sent the next frame.

The idle-RQ protocol operates in a half duplex mode and is typically used in

master slave environments. So the master sends a frame and must wait for the

response of the slave whether the frame was properly received or not.

In today's networks the idle-RQ technique is seen very rarely, because it

introduces large delays and is not able to fill data pipes we are currently used

to.

23


Without Sequence Numbers:

Sender Receiver

Data "ABCD"

Ack

Ack

Ack

Data "ABCD"

Data "EFGH"

No Ack?

Retransmission!

ABCD

ABCD

ABCD

ABCD

ABCD

EFGH

There are two different ways how idle-RQ technique might be implemented.

In this graphic an idle-RQ system without the use of sequence numbers is

shown. The sender sends out a data frame, starts an retransmission timer and

waits for the receive of an acknowledgement.

An already sent data frame remains in the send buffer and may only be deleted

if an proper acknowledgement is received.

A data frame retransmission will happen, if the retransmission timer times out

before an acknowledgement is received. Even if the transmission itself was

successful and only the acknowledgement was lost. This could lead to an

transport of duplicate data frames.

24


With Sequence Numbers:

Sender Receiver

Data "ABCD" S=0

Ack=0

Ack=0

Ack=1

Data "ABCD" S=0

Data "EFGH" S=1

No Ack?

Retransmission!

ABCD

ABCD

ABCD

EFGH

In the second scenario of idle-RQ systems sequence numbers are used.

In this case the sender sends out a data frame with a valid sequence number.

Keeps the data frame in the send buffer and starts the retransmission timer. The

acknowledgement frame is lost again and the data frame is retransmitted.

But now the receiver is able to recognize, by the help of the sequence number,

that the received data frame is a duplicate and is able to discard it.

25


Slow !

Vienna Tokyo

Data

Ack

"Stop and Wait":"Stop and Wait":

Data is traveling round

the earth while sender

waits for the

acknowledgement.

In the meantime

no data can be sent !!

Data

As already mentioned before the idle-RQ technique is not suited for today's

data transport systems. The stop and wait procedure would introduce a lot of

delay, especially on long distance connections, and would no be able to

efficiently use the network infrastructure.

26


Empty Pipe !

Vienna Tokyo

t = 0 s

t = 350 ms

1 KB Data

Ack

1.5 Mbit/s

This pipe allows

1.5 Mbit/s 350 ms

= 525,000 bit @ 64 KB

of data.

But stop-and-wait only

allows one frame to be

outstanding !!!

Assume MTU=1 KByte,

then the max rate is

(1024 8) bit / 0.35 s

@ 23 Kbit/s

In this scenario we have a 1.5 Mbit/s connection between Vienna and Tokyo.

A 1 KByte data frame is sent from Vienna to Tokyo with a transport delay of

175 milliseconds in one direction. With the use of idle-RQ technique it would

take at least 350 milliseconds before an acknowledgement is received and the

next data frame is sent.

In this case a maximum throughput of only 23 Kbit/s can be achieved.

27


Idle-RQ Facts

Old and slow method

But small code and only little resources

necessary

At least two sequence numbers

necessary

To distinguish new from old data

Half duplex protocol

Example: TFTP

The only advantage of the idle-RQ technique compared to the more

sophisticated continuous RQ techniques is the little amount of memory and

processor resources that is needed.

Therefore it is very easy to implement them in ROM based systems e.g. cisco

router support TFTP from ROM monitor

28


Continuous RQ

Data and Acks

are sent

continuously !!

Data

Ack

Data

Data

Data

Ack

Ack

Ack

Acks

DataIdle-RQ Continuous-RQ

In continuous-RQ technology data frames and their according

acknowledgements are sent continuously.

The sender is allowed to put a certain amount of frames into the send buffer

and transmit them all in one go. The amount of frames the sender is allowed to

send is either negotiated during the connection establishment phase or

dynamically adjusted by max window size announcements of the received

acknowledgements.

29


Full Pipe !

Vienna Tokyo

t = 0 s

t = 350 ms

1 KB Data

Ack

1.5 Mbit/s

This situation

corresponds with a

sliding window

By the use of continous-RQ technique and an proper adjusted send window

size it would now be possible to use the complete capacity of our Vienna

Tokyo connection.

30


Need For Retransmission Buffer

Vienna Tokyo

Data S=0Data S=1Data S=2Data S=3

Four packets are sent,

but due to a network

failure none of them

arrive (or equivalently

they do arrive but the

Acks are lost)

Timeouts !!!

0

0 1

0 1 2

0 1 2 3

Data S=0Data S=1

.

.

.

.

0 1 2 3

0 1 2 3

0 1 2 3

0 1 2 3

Retransmissions

Ack

0

1 0Ac

k

In this example four data frames are sent from Vienna to Tokyo and all of them

are lost or the according acknow

Data Communication Fundamentals

Documents

Transcript of Data Communication Fundamentals