Asynchronous Transfer Mode (ATM)

Definition
Asynchronous refers to the manner in which BW is allocated
among connections and users. ATM is asynchronous in the sense that
the recurrence of cells containing information from an individual
user is not necessarily periodic.

ATM uses statistical (asynchronous) time-division multiplexing i.e.

the multiplexer fills a slot with a cell from any input channel that has
a cell; the slot is empty if none of the channels has a cell to send.

Transfer mode is a term intended to signify that it is a

multiplexing and switching technique.
1
 ATM Multiplexing
Fig. below shows how cells from three inputs are multiplexed. At
the first tick of the clock, channel 2 has no cell (empty input slot),
so multiples fills the slot with a cell from third channel. When all
the cells from all the channels are multiplexed, the output slots are
empty.

Fig.1 ATM 2
multiplexing
 Asynchronous Transfer Mode (ATM)

Voice
Data
packets MUX

Wasted bandwidth
Images

TDM
4 3 2 1 4 3 2 1 4 3 2 1

ATM `
4 3 1 3 2 2 1

3
ATM is a packet-oriented (also connection oriented) transfer mode
which allows multiple logical connections to be multiplexed over a
single physical interface. The information flow on each logical
connection is organized into fixed-size packets called cell.

Why small cell?

It is easy to build hardware routers to handle short, fixed-length
cells. Variable length IP packets have to be routed by software,
which is slower process.
Hardware can copy one incoming cell to multiple output lines
easily which is suitable for TV broadcasting.
Small cell can not block a line for long which ensure QoS.
4
Applications of ATM
To support any type of traffic:
- burtsy data (to multimegabit rates: files, images, multimedia)
- voice (sustained data rate, 64 kbps)
- video (sustained data rate, multimegabit rates)
To support transactions that use data, voice, and video simultaneously
To provide bandwidth on demand (pay for use)
To support multicast operations (video conferencing)
To provide guaranteed quality of service

5
 Architecture of an ATM Network

The user access devices, called the end points are connected through
a user-to-network interface (UNI) to the switches inside the network.
The switches are connected through network-to-network interface
(NNI)

6
 Virtual Connection
There exist two types of ATM logical
connections:
•Virtual Path Connections (VPs)
•Virtual Channel Connections (VCs).
A transmission path (TP) is the physical connection (wire, cable, satellite etc)
between and end point and switch or between two switches. A transmission path can
be considered as the set of all highways that directly connect two cities. A
transmission path is divided into several virtual paths (VPs). Think virtual path as a
highway between two cities. Each lane of the highway (virtual path) is considered as
virtual circuit (VCs).
Virtual Paths
Physical Link (TP)

Virtual Channels

7
 Identifiers
Each connection is identified by a Virtual Path Identifier (VPI)/Virtual
Circuit Identifier (VCI) combination, which can be found in the header
of all ATM cells carrying information about that specific connection.
A group of VC links, identified by a common value of VPI.

8
ATM Switching

9
A Conceptual View of a VP Switch

10
A Conceptual View of a VP and VC Switch

11
12
13
14
 ATM Protocol Hierarchy
ATM has its own reference model different from OSI or TCP/IP
model. The ATM standard defines three layers. They are:
application adaptation layer, the ATM layer and physical layer.

Upper Layers
Convergence Sublayer
• Provide application-specific interface
• Handle lost and delayed cells
• Error detection and handling

Segmentation and Reassembly Sublayer

• Pack Convergence Sublayer information into 48-byte
blocks for transfer down to the ATM Layer.
• Unpack ATM Layer cells for transfer up to the
Convergence Sublayer.

ATM Layer
Physical Layer 15
Physical layer
The physical layer is responsible for transmission of ATM cells as a
bitstream across the physical medium like Ethernet or wireless LAN.
ATM layer
The ATM later provides routing, traffic management, switching and
multiplexing services. It processes outgoing traffic by accepting 48 byte
segments from AAL and transforming them into 53 bytes cells by
addition of 5-byte header.
Application Adaption Layer
The use of ATM creates the need for an adaptation layer to support
information transfer protocols not based on ATM. For example PCM
voice is an application that produces a stream of bits from a voice signal.
To employ this application over ATM, it is necessary to assemble PCM
bits into cells for transmission and to read them out on reception in such
a way to produce a smooth, constant flow of bits to the receiver. 16
 Application Adaption Layer
Voice

A/D AAL
s1 , s2 … cells
Digital voice in PCM

Video

A/D … Compression AAL

cells
compressed
picture frames
frames

Data AAL
Bursty variable-length cells
packets
17
 AAL user The AAL layer is organized in to logical
sublayers: The convergence sublayer makes it
CS possible to have ATM systems offer different
AAL kinds of services of different applications (like file
SAR sublayer transfer, video on demand etc).

ATM layer
The segmentation and reassembly sublayer is
responsible to break up the message (from CS
Physical layer
sublayer) into cells on transmission side and puts
them back together again at the destination side.

18
 ATM and AAL connections
end-system A end-system B

service dependent
AAL AAL connections
AAL

service independent
ATM ATM
ATM connections
physical physical
layer layer

ATM network

ATM layer:
service independent transport of ATM cells application
multiplex and demultiplex functionality
AAL layer: support of different services
19
 Function of layers on transmitting side
Application Layer

Message

Convergence Sublayer

CS CS
Header Trailer

Segmentation and Reassembly Sublayer

Pad

Segmentation and Reassembly Sublayer (continued)

SAR SAR SAR SAR SAR SAR SAR SAR SAR SAR
Hdr Trlr Hdr Trlr Hdr Trlr Hdr Trlr Hdr Trlr

ATM Layer
ATM ATM ATM ATM ATM
Hdr Hdr Hdr Hdr Hdr

20
ATM Cell Format

21
 ATM Cells
The asynchronous transfer mode makes use of fixed-size cells,
consisting of 5-octat header and 48-octet information field

Basic ATM Cell Format

5 Bytes 48 Bytes

Header Payload

Use of small cells reduce queuing delay for high priority

cell.
Fixed-size cells can be switched more efficiently.
22
ATM Cell Format
Generic Flow Control:
relieves short-term congestion
problems between the user and
the ATM network

GFC VPI GFC Virtual Path Identifier:

VPI VCI 8-bit routing field in UNI; 12-bit
routing field in NNI (to support
VCI larger number of internal paths)
VCI PT C VPI
Virtual Channel Identifier:
HEC 16-bit routing field to
and from end users
Payload VCI
Payload Payload Type:
Type of info in cell (e.g., user
traffic, control, maintenance)
48-Bytes PT
Cell Loss Priority:
Total signifies whether sender
Payload exceeded negotiated
C transmission rate

Payload Header Error Check:

CRC for checking errors and
HEC synchronization
Payload
23
24
GFC VPI VCI PT CLP HEC Payload
4bits 8bits 16bits 3bits 1bits 8bits
48bytes
UNI Cell
VPI VCI PT CLP HEC Payload
12bits 16bits 3bits 1bits 8bits
48bytes

NNI Cell

The CLP bit allows a two level loss priority (CLP = 0 for high,
CLP = 1 for low) be specified explicitly for individual cells. Low
loss priority cells should be discarded before cells of high loss
priority if congestion occurs and necessities a loss of cells.
HEC uses CRC code
25
 User User
information information

AAL AAL

ATM ATM ATM ATM

PHY PHY PHY PHY

…
End system Network End system

26
Request for Call establishment using virtual path
VCC originates

Yes Can QoS be Yes

VPC exists?
satisfied?
No
No
Block VCC or
Establish request more
capacity
new VPC

Make
Request Yes
connection
granted?

No
Reject VCC
request

27
 Modeling of M/D/1 asynchronous
traffic (ATM system)
Most of the traffics in telecommunications network follow
exponential arrival and exponentional service time distribution
like M/M/n/K/N as complete notation. In ATM network service
time of each cell/packet is fixed hence deterministic service time
traffic like M/D/n/K/N is applicable to detect traffic parameters for
that network. In this section a traffic model of single server case,
where arrival of traffic is Markovian but the service time is
deterministic i.e. M/D/1 traffic.

28
Let,
Pt(x) =P{The system in state x at time t}
ai = P{during time interval (t, t+tm) i requests arrive}
tm= mean holding time=1 time unit
A=tm ; is the mean of Poisson’s offered traffic
To reach the state x(t+1) from x(t) transitions take place shown
in table-1.

29
State x(t) Events(s) P{events(s)} State x(t+1)
0 No arrival a0 0
0 1 arrival a1 1
0 x>1 arrival ax>1 x>1

1 No arrival a0 0
1 1 arrival a1 1
1 2 arrivals a2 2
1 x>2 arrivals x>2
ax>1 Table-1
2 No arrival a0 1
2 1 arrival a1 2
2 2 arrivals a2 3
2 x>2 arrivals 1+x>2
ax>1
3 No arrival a0 2
3 1 arrival a1 3
3 2 arrivals a2 4
3 x>2 arrivals 2+x>2
ax>1
x No arrival a0 x-1
x 1 arrival a1 x 30
x 2 arrivals a2 x+1
Balanced equations from above table could be detected as,
Pt+1(0)=Pt(0)a0+Pt(1)a0
Pt+1(1)=Pt(0)a1+Pt(1)a1+ Pt(2)a0
Pt+1(2)=Pt(1)a2+Pt(2)a1+ Pt(3)a0 etc
In statistical equilibrium state actual point (t) on time axis
does not have any influence on the probabilities, become
time-independent steady state probabilities P(x)=Px
P0=P0a0+P1a0
P1=P0a1+P1a1+ P2a0
P2=P1a2+P2a1+ P3a0 etc.
These equations could be solved recursively say,

31
 P0 (1  a 0 )
P1 
a0
 P0 a 1  P1 ( 1  a 1 )
P2 
a0
Here ai follows Poisson’s process,

ai 
  t m   t
i
e m

 tm !

32
 Since buffer is infinite hence carries traffic,


P
x 1
x  1  P0  A

 P0  1  A

P0 (1  e A )
P1  A
 P0 ( e A
 1)
e
 
P2  e A  P0 Ae  A  P1 ( 1  Ae A )  ( 1  A )e A [ e A  1  A ]

………………………………………………………..
………………………………………………………..
x  ( kA ) x  k ( kA ) x  k 1 
Px  ( 1  A )  ( 1 )
k 1
xk
e kA
 
 ( x  k )! ( x  k  1 )! 
;x  2

33
The second factor inside the parenthesis is ignored for k=x.
Here A<1, there is no call blocking and the carried traffic is,


X  1  P0  A   Px
x 1

 P0  1  A

34
The asynchronous M/D/1 formulas are well valid for a synchronous
M/D/1 queue with internal server (S-IS M/D/1). In this system
during the time slot ( cell time in ATM system) cell arrive
according to Poisson process, which is defined by a single
parameters, the mean arrival rte λ.
ATM Queue

Group of 5 4 3 2 Departure λ(1-B)

Users Average
arrival rate λ

Time slots of length tm

K=∞ Queue/ Buffer

35
 Queue

Departure-1 Departure-2 Departure-3

Cell-5 No arrival in these
slots
Cell-4 Cell-5
4 Arrivals in
this TS
Cell-3 Cell-4 Cell-5

Cell-1 Cell-2 Cell-3 Cell-4

t t+tm t+2tm t+3tm

At the end of the TS the queue (internal server) first removes a cell (if any) from
the queue (its service time is complete) and then stores the new cells. Each cell
spends in the queue a minimum time of 1 slots (cell service time). Mean waiting
time is same as in normal M/D/1.
For ‘integer’ waiting times that are integer multiples of service time the waiting
time distribution is equal to the system sate distribution. Let P{tw≤k.tm) be the
probability that waiting time is less or equal than k service times and Px is sate
probability, then,
36
P{tw ≤ k.tm)=P0+P1+P2+… … … +Pk

k
P{t w  k}   Pi
i 0

37