Chapter3

1

Connection-Oriented Networks - Harry Perros 1

Chapter 3:ATM Networks

TOPICS– The ATM header– The ATM protocol stack– The physical layer– ATM switch architectures– ATM adaptation layers– IP over ATM


Asynchronous Transfer Mode (ATM)

• The word Asynchronous in ATM is incontrast to Synchronous Transfer Mode(STM) that was proposed earlier on, whichwas based on the SONET/SDH hierarchy.

• Transfer Mode refers to a telecommunicationtechnique

2


• ATM was standardized by ITU-T (old CCITT) in1988 as the transfer mode of B-ISDN

• It can carry a variety of different types of traffic,such as– Voice– Video– Data

At speeds varying from fractional T1 to 2.4 Gbps


• These different types of traffic havedifferent Quality-of-Service (QoS)requirements, such as:– Packet loss– End-to-end delay

• ATM, unlike IP networks, can provide eachtraffic connection a different type of qualityof service.

3


Some features of ATM

• Connection-oriented packet-switched network• Fixed cell (packet) size of 48+5 bytes

• No error protection on a link-by-link• No flow control on a link-by-link• Delivers cells in the order in which they were

transmitted

Header Payload

5 bytes 48 bytes


The structure of the ATM cell

GFC VPI

VPI VCI

VCI

VCI

HEC

PTI CLP

Informationpayload

1 2 3 4 5 6 7 81

2

3

4

5

.

.

.

53

Byte

UNI cell format

VPI

VPI VCI

VCI

VCI

HEC

PTI CLP

Informationpayload

1 2 3 4 5 6 7 81

2

3

4

5

.

.

.

53

Byte

NNI cell format

4


Fields in the ATM cell header

• GFC• Connection identifier: VPI/VCI,• Payload type indicator (PTI)• Cell loss priority (CLP)• Head error control (HEC)


ATM connections

• Identified by the combined fields– virtual path identification (VPI), and– virtual channel identification (VCI)

• VPI field:– 256 virtual paths at the UNI interface, and – 4096 virtual paths at the NNI interface.

• VCI field:– a maximum of 65,536 VCIs.

5


• VPI/VCI values have local significance.That is, they are only valid for a single hop.

• A connection over many hops, is associatedwith a different VPI/VCI value on each hop.

• Each switch maintains a switching table.For each connection, it keeps the incomingand outgoing VPI/VCI values and the inputand output ports.


Label swapping

VPI=10VCI=89

VPI=30VCI=53

VPI=100VCI=53

VPI=50VCI=77

VPI=30VCI=41

30 53 4 100 53 5

30 41 1 30 53 440 62 2 10 89 3

10 89 1 50 77 6

ATMswitch 1

ATMswitch 2

ATMswitch 3

VPI=40VCI=62

A

B

C

D2

1

3

4 4

1

5

6

6


PVCs and SVCs• Depending how a connection is set-up, it

may be– Permanent virtual circuit (PVC)– Switched Virtual circuit (SVC)

• PVCs are set-up administratively. Theyremain up for a long time.

• SVCs are set-up in real-time using ATMsignalling. Their duration is arbitrary.


Payload type Indicator

• PTI Meaning • 000 User data cell, congestion not experienced, SDU type=0• 001 User data cell, congestion not experienced, SDU type=1• 010 User data cell, congestion experienced, SDU type=0• 011 User data cell, congestion experienced, SDU type=1• 100 Segment OAM flow-related cell• 101 End-to-end OAM flow-related cell• 110 RM cell• 111 Reserved

7


No error detected(No action)

Error detected(cell discard)

Head Error Control (HEC)

Correctionmode

DetectionmodeNo error detected

Single bit error detected(correction)

Multiple bit error detected(cell discarded


The ATM protocol stack

ATM adaptation layer

ATM layer

Physical layer

voice Video Data

8


The physical layer

• The physical layer transports ATM cellsbetween two adjacent ATM layers.

• It is subdivided into– transmission convergence (TC) sublayer– physical medium-dependent (PMD) sublayer.


The transmission convergence (TC)sublayer

• HEC cell generation and verification– Implements the HEC state machine

• Decoupling of cell rate– Maintains a continuous bit stream by inserting idle cells

• Transmission frame generation and recovery– Such as SONET frames

• Cell delineation

9


Cell delineation is the extraction of cells from thebit stream received from the PMD sublayer.

hunt

Presync

Sync

CorrectHEC

CorrectHEC for δ cells

IncorrectHEC for α cells

IncorrectHEC


• Physical medium dependent (PMD)– Timing function

• Used to synchronize the transmitting and receivingPMD sublayers.

– Encoding/decoding• PMD may operate on a bit-by-bit basis or using

block coding such as 4B/5B and 8B/10B schemes.

10


ATM physical layer interfaces

• SONET/SDH• Plesiochronous digital hierarchy (PDH)• Nx64 Kbps• Inverse mulitplexing for ATM (IMA)• asymmetric digital subscriber line (ADSL)• APON


The ATM layer

• The ATM layer is concerned with the end-to-end transfer of information, i.e., from thetransmitting end-device to the receivingend-device.

• Below, we summarize its main features.

11


Connection-oriented packet switching

• The ATM layer is a connection-oriented point-topoint packet-switched network with fixed-sizepackets (known as cells).

• A connection is identified by a series of VPI/VCIlabels, as explained above, and it may be point-to-point or point-to-multipoint.

• Cells are delivered to the destination in the orderin which they were transmitted.


Cell switching in ATM networksis carried out at the ATM level

End-device

Application Application

ATM switch End-deviceATM switch

12


No error and flow controlon each hop

• Low probability of a cell getting lost or delivered to thedestination end-device in error.

• The recovery of the data carried by lost or corrupted cellsis expected to be carried out by a higher-level protocol,such as TCP.

• When TCP/IP runs over ATM, the loss or corruption of thepayload of a single cell results in the retransmission of anentire TCP PDU.


Addressing

• Each ATM end-device and ATM switch has a uniqueATM address.

• Private and public networks use different ATMaddresses. Public networks use E.164 addresses andprivate networks use the OSI NSAP format.

• ATM addresses are different to IP addresses.

13


Quality of service

• Each ATM connection is associated with aquality-of-service category.

• Each quality-of-service category is associated witha set of traffic parameters and a set of quality-of-service parameters.

• The ATM network guarantees the negotiatedquality-of-service for each connection.


Congestion control

• In ATM networks, congestion controlpermits the network operator to carry asmuch traffic as possible without affectingthe quality of service requested by the users.

• It consists of call admission control and apolicing mechanism.

14


A generic ATM switch

Incominglinks

. . .

. . . Outgoinglinks

Inputqueues

Outputqueues

CPU

Switch fabric

1

N N

1

The ATM switch architecture


The shared memory ATM switcharchitecture

. . .

. . .

Shared memory1

N N

1

15


• If the speed of transmission on eachincoming and outgoing link is V, then theswitch can keep up at maximum arrival rate,if the memory's bandwidth is 2NV

• Total memory capacity is B cells• Each linked list i is associated with a

minimum dedicated space and it is limitedto an upper bound Bi, Bi<B so that ΣBi>B.


Cell loss in a shared memory switch

• Cell loss occurs when a cell arrives at a time whenthe shared memory is full, that is it, contains Bcells.

• Cell loss can also occur when a cell withdestination output port i arrives at a time when thetotal number of cells queued for this output port isBi cells (even if the total number of cells in theshared memory is less than B.)

16


Non-blocking output buffering switch• In a non-blocking switch, the switching fabric

does not give rise to internal or external blocking.• An output buffering switch has buffers only at its

output ports.

. . .

. . .

output ports

• A shared memory switch is non-blocking withoutput buffering


Scheduling algorithms

• Let us consider a non-blocking switch with outputbuffering. Each output buffer holds cells thatbelong to different connections.

• Each of these connections is associated with aquality-of-service category.

• The cells belonging to these connections aregrouped into queues, one per quality-of-servicecategory, and these queues are served using ascheduling algorithm.

17


Static priorities

From switchfabric

CBR

RT-VBR

NRT-VBR

ABR

UBR

Output port

• Priorities among thequeues

• Always serve highpriority queue first,then second priorityqueue, etc.

• Aging• Purging


Early deadline first (EDF) algorithm

• Each cell is assigned a deadline upon arrival atthe buffer. The scheduler servers the cellsaccording to their deadlines, so that the one withthe earliest deadline gets served first.

• Cells belonging to delay-sensitive applications,such as voice or video, can be served first byassigning them deadlines closer to their arrivaltimes.

18


The round-robin scheduler

• Each output buffer is organized into a number oflogical queues.

• The scheduler serves one cell from each queue ina round robin fashion

• Empty queues are skipped• Weighted round robin scheduling can be used to

serve a different number of cells from each queue


The ATM adaptation layer

• The purpose of AAL is to isolate higherlayers from the specific characteristics ofthe ATM layer.

• AAL consists of the– convergence sublayer, and the– segmentation-and-reassembly sublayer.

19


The AAL sublayers

SAP

Service SpecificConvergence Sublayer (SSCS)

Common Part Sublayer (CPS)

Segmentation and Reassembly

Convergence Sublayer

SAP


ATM Adaptation Layer 1 (AAL 1)

• This AAL can be used for applications such as:– circuit emulation services

• It emulates a point-to-point TDM circuit over ATM– Constant-bit rate audio

• Used to provide an interconnection between twoPBXs over a private or public ATM network

20


The SAR encapsulation for AAL 1

payloadSAR Header

47 bytesSN SNP

CRC-3

3 bits

Parity

1 bit3 bits1 bit

CSI Sequence. count


The AAL 1 CS functions:1. Handling of cell variation

– Due to queueing delays, inter-arrival times of cells vary (jitter).

– CS writes received data into a buffer, and then delivers theinformation to the application at constant bit rate.

ATM cloud

Inter-departure gaps

Sender

cell i-1

Inter-arrival gaps

Receiver

cell i

celli+1

cell i-1

cell i

celli+1

t i-1 t i si-1 si

21


2. Processing of the sequence count– The sequence count values are processed by CS in

order to detect lost or misinserted cells. Detectedmisinserted cells are discarded. In order to maintainbit count integrity of the AAL user information, itmay be necessary to compensate for lost cells byinserting dummy SAR-PDU payloads.

3. Forward error correction– For video and high quality audio forward error

correction may be performed in order to protectagainst bit errors. This may be combined withinterleaving of AAL user bits to give a more secureprotection against errors.


4. Transfer of timing informationa. Synchronous residual time stamp (SRTS):

CS conveys to the receiver in the CSI field thedifference between a common clock derived from thenetwork and the sender’s clock

b. Adaptive clock method:The receiver writes the received information into abuffer and reads out from the buffer. If its clock isfast/slow the occupancy in the buffer will bebelow/over the median

22


5. Structured and unstructured data transfersTwo CS-PDU formats have been defined:a.CS-PDU non-P format:

Constructed from 47 bytes of informationsupplied by an AAL user

b. CS-PDU P format:Constructed from a 1-byte header and 46 bytesof information supplied by an AAL use.The header consists of a 7-bit pointer (SDTpointer) and 1 even bit parity.


Circuit Emulation Services

IWFA

IWFBATM network

UNI UNI

CBRUser A

CBRUser B

• The structured and unstructured data transfers areused in Circuit Emulation Services (CES), whichemulate a T1/E1 link over ATM.

• CES is implemented in an interworking function(IWF).

23


• Unstructured service– The entire DS1/E1 signal is transported by

packing it bit by bit into the 47-byte payload ofa CS-PDU non-P format, which is then carriedby an ATM cell.

47 bytes -> 376 bits -> less than 2 DS-1 frames (193 bits/frame)


• Structured transfers– It is used to carry fractional T1/E1, i.e. Nx64

Kbps– Fractional T1/E1 generates blocks of N bytes

every 125 µsec. Such a block of data is referredto in the standards as a structured block.

– Blocks of N bytes are transported back-to-backover successive cells using both the CS-PDUnon-P and P formats.

24


• The SDT pointer– The SDT pointer in the CS-PDU P format is

used to help delineate the boundaries of theseblocks.

– The actual rules as to when to use the SDTpointer in the P format are somewhat complex.


An example: block size = 20 bytes

Seq, count 0 Seq, count 1 Seq, count 2 Seq, count 3



20 20 6 14 20 13 7 20 20 20 20 7

13 20 14 6 20 20 19 20 7 13 20 14

6 20 20 20 20 7 13 20 14 6 20 20 1

1

25


ATM Adaptation Layer 2 (AAL 2)

• Defined for delay sensitive applicationswith a low bit rate, such as voice andvoiceband traffic (facsimile, modem traffic)

• AAL 2 is used in to interconnect two distantpublic or private telephone networks overan ATM network.


• At the sender, AAL 2 multiplexes several streamsonto the same ATM connection

• At the receiver, it de-multiplexes the date fromthe connection to the individual streams.

26


The SSCS and CPS sublayers

• The AAL 2 services are provided by theconvergence sublayer, which is subdivided intothe– Service Specific Convergence Sublayer (SSCS)– Common part sublayer (CPS).


Functional model of AAL 2 (sender side)

• Each stream is served by a separate SSCS which isassociated with a CID

SSCS

AAL-SAP

ATM-SAP

SSCSSSCS

CID=XCID=Y

CID=Z

CPS

27


SSCS for AAL 2 trunking

• A specialized SSCS has been developed tosupport “ATM trunking using AAL 2 fornarrowband services”.

• It is described in Chapter 12


• A transmitting SSCS uses a timer to decidewhen to pass on the data to CPS.

• Data from an SSCS is packed into a CPS-packet• CPS-packets from different SSCSs are packed

into a CPS-PDU, which is exactly 48 bytes andit is carried in an ATM cell

CPS-packets and CPS-PDUs

28


Packing CPS-packets into CPS-PDUs

CPS-PDUs

1 2 3 4 5

54321 3

ATM cells

CPS-packets


The structure of the CPS-packetand CPS-PDU

OSFP

CPS-PDUpayload

SN

PAD

C I D

PPT LI

HEC UUI

CPS-packetpayload

1

2

3

1

48

CPS-packet CPS-PDU

1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8

29


The CPS-packet fields• Channel identifier (CID) - 8 bits:

– Identifies a channel. Same value is used for both directions.– CIDs are allocated using the AAL negotiation procedures (ANP)

• Packet payload type (PPT) - 2 bits:– Indicates whether it carries voice or network management data

• Length indicator (LI) - 6 bits:– Default maximum length of the CPS-packet payload is 45 bytes.

• Header error control (HEC) - 5 bits:– Pattern is: x5+x2+1.

• User-to-user-indication (UUI) - 3 bits:– Used to transfer information transparently between the peers.


CPS-PDU fields

Offset field (OSF) (6 bits)• Used to identify the beginning of a CPS-

packet. It points to the first new CPS-packetin the CPS-PDU payload

• In the absence of a new CPS-packet, itpoints to the beginning of the pad

• The value of 47 indicates that there is nobeginning of a CPS-packet in the CPS-PDU.

30


An example

20 48 35 20

OSF =0 OSF=21 OSF=9

20 27 21 26 9 20 #1#2

#2 #3 #3 #4padding


ATM adaptation layer 5 (AAL 5)

• Very popular AAL due to its simplicity• A user-PDU is encapsulated and then broken up to

fragments, each carried by an ATM cell• AAL 5 consists of

– Convergence sublayer (CS)• SSCS• CPS

– Segmentation and reassembly (SAR).

31


• Provides a non-assured transfer operation.• User-PDUs of a length up to 65,535 bytes

can be transferred.• Erroneous CPS-PDUs can be detected at the

receiver’s side. No recovery of an erroneousCS-PDU takes place. Instead, an indicationis sent to the the higher-level application.

CPS


• Pad: from 0 to 47 bytes, so that the entire CPS-PDU becomesan integer multiple of 48 bytes. The User-PDU can be up to65,535 bytes

• CPS User-to-user indication (CPS-UU): 1-byte field• Common part indicator (CPI): 1-byte field for future use• Length: 2-byte field gives length of CPS-PDU payload.• CRC pattern: 4-byte field contains the FCS calculated using

the pattern x32+x26+x23+x22+x16+x12+x11+x10+x8+x7+x5+x4+x2+x+1.

User-PDU Pad CRC-32LengthCPS-UU

CPS encapsulation

32


SAR sublayer - transmitter

• SAR segments a CPS-PDU into a sequenceof 48-byte segments.

• No additional encapsulation• Each segment is carried in the payload of an

ATM cell• Last cell has its PTI marked with SDU=1.


SAR sublayer - Receiver

• SAR appends payloads of the ATM cellinto a buffer until1. It encounters an SDU=1 in PTI field

• It checks the CRC and then passes the PDU to theapplication above with an indication as to whetherit is correct or not.

2. Buffer is exceeded• It passed the PDU to the application above with an

indication that buffer was exceeded.

33


Classical IP and ARP over ATM

• This is a technique proposed by IETF forsupporting IP over ATM in a single logical IPsubnet (LIS).

• A LIS is a group of IP hosts that share a commonIP network number and subnet mask, and they allcommunicate with each other directly over ATMconnections.


Group of host with the same network addressand same subnet maskIP address: 193.14.0.0

We now replace the transport network with ATM switches

A logical IP subnet (LIS)

34


PHYATMAALCIPIP

TCP

PHYATMAALCIPIP

TCP

IP packet

CS-PDU

SAR

ATM

IP packet

CS-PDU

SAR

ATM

Computer Computer

Classical IP

ATM switch


ATMARP and InATMARP

• A LIS member has both an IP and an ATMaddress.

• IP addresses are resolved to ATM addresses usingthe ATMARP protocol within the LIS (based onARP).

• The inverse ATMARP (InATMARP) protocol isused to resolve an ATM address to an IP address.(It is similar to the inverseARP).

35


The ATMARP server

• It maintains a table or a cache of IP and ATMaddress mappings

• It learns about the IP and ATM addresses of theLIS members (IP clients) through the messagesexchanged between ATMARP and the LIS hosts.

• It typically resides on an ATM switch. (An ATMswitch load commonly contains the ATMARPserver as well).


ATMARP messages

• ATMARP_request: Used to request the ATMARP serverthe ATM address of a destination IP client.

• ATMARP_reply: Used by the ATMARP server to respondto an IP client with the destination ATM address.

• InATMARP_request: Sent from the ATMARP server to anIP client to obtain its IP address.

• InATMARP_reply: This is the response from an IP clientwith its IP address.

• ATMARP_NAK: Negative response issued by theATMARP server to a requesting IP client.

36


Registration

• An IP client must first register its IP and ATMaddresses with the ATMARP server.– The client establishes a connection to the ATMARP

server (it knows the ATMARP server’s ATM address).– It then transmits an ATMARP_request, asking its own

ATM address.– The ATMARP server checks against duplicate entries

in its table, time stamps the entry, adds it to its table,and sends an ATMARP_reply. The entry is valid for aminimum of 20 minutes and has to be refreshed.


Address resolution• An ATMARP client 1 can communicate with an

ATMARP client 2 immediately if it knows itsATM address.

• If the destination ATM address is not known,client 1 invokes the ATMARP process.– It sends an ATMARP_request to the ATMARP server.– If the server has the requested address in its table, it

returns an ATMARP_reply.– Otherwise, it returns an ATMARP_NAK.

Education

Chapter3