2. B-ISDN REFERENCE MODEL and

PROTOCOL LAYERS

Management Plane

AAL

ATM Layer

Physical Layer

Signaling Protocol

Applications

TCP/IPNative

User Plane Control Plane

B-ISDN Protocol Reference Model

SNMP: Simple Network Management Protocol

CMIP: CommonManagementInformationProtocol

Control Plane Supports Signaling Call Setup, Call Control, Connection Control

User Plane Data Transfer, Flow Control, Error Recovery

Management Plane Operation, Administration, & Maintenance

Management Plane(Provides Control of ATM Switch)

Layer Management(Layered)

Plane Management(No Layered)

– Use to manage each of the ATM layers with entity corresponding to each ATM layer

– OAM issues

– Concerned with management of all the planes

– All management functions (Fault, Performance, Configuration, Operation, & Security) which relates to the whole system are located in the Plane Management

– Provides coordination between all planes

Broadband Networking with SONET and ATM

USER

USER

USER

USERATM SW ATM SW

VideoImageDataetc…

UNI UNINNI

Higher Layers

ConvergenceSublayers (CS)

SegmentationReassembly

Sublayer (SAR)

ATM Layer

Physical Layer

•Flow Control•Error Handling•Message Segmentation

•Segmentation Type•Message Number•Message ID

•5 Byte Header•48 Byte Payload•Handles cont. and bursty traffic

•SONET

Higher Layers

ConvergenceSublayers (CS)

SegmentationReassembly

Sublayer (SAR)

ATM Layer

Physical Layer

USER USER

AdaptationLayer

ATM Layer

Physical Layer

ATM Layer

Physical Layer

Protocol Reference Model in the User Plane

Upper Layers

1 2 3 4class A class B class C class D

CellInformationFieldAAL

ATM

PL

CS

SAR

•Handling lost / misdelivered cells•Timing recovery•Interleaving

•Split frames / bit stream info cells•Re-assemble frames / bit stream

•Cell routing•Multiplexing / demultiplexing•Generic flow control

TC

PM

•Cell header verification and cell delineation•Rate decoupling (insert idle cells)•Transmission frame adaptation•Bit timing•Physical medium

CellHeader

AAL = ATM Adaptation LayerSAR = Segmentation and ReassemblyCS = Convergence SublayerPL = Physical LayerTC = Transmission ConvergencePM = Physical Medium

Abbreviations

Class

Service Classes for AALType

ABCD

Constant Bit RateVariable Bit RateConnection Oriented DataConnectionless Data

• SEAL = Simple and Efficient Adaptation Layer•Type 5 AAL•Acknowledged info transfer

Remark: See next page

Remarks: PMD Physical Medium Dependent

TC Transmission Convergence

Sublayer

It separates transmission from the physical interface and allows ATM interfaces to be built

on a large variety of physical interfaces

a) Physical Medium (PM)– PM sublayer provides the bit transmission

capability including bit alignment– Line coding and, if necessary,

electrical/optical conversion is performed in this sublayer

– Optical fiber is used for the physical medium. Other media, coax cables are also possible

– Bit rates 155 Mbps or 622.080 Mbps.

Physical Layer Functions

b) Bit Timing– Generation and reception of waveforms

which are suitable for the medium, the insertion, and extraction of bit timing information and the line coding if required

– CMI (Code Mark Inversion) (CCITT G.703) proposed for 155.520 Mbps interface.

– NRZ “Nonreturn to Zero” code proposed for optical interface.

PHYSICAL LAYER FUNCTIONS

Electrical Interface: Coded Mark Inversion (CMI)– For binary 0 always a positive transition at the

midpoint of the binary unit time interval.– For binary 1 always a constant signal level for the

duration of the bit time. This level alternates between high and low for successive binary 1s.

0 0 0 0 0 01 1 1

Level A1

Level A2

1 1

LINE CODING

Optical Interface: Nonreturn to Zero (NRZ)

– For binary 0 Emission of light– For binary 1 No emission of light– Transition: 0 1 or 1 0

Otherwise no transition

0 0 0 0 0 01 1 1

Level A1

Level A2

1 1

LINE CODING

• SONET/SDH : 155 Mbps and 622 Mbps over OC-3 (single mode fiber)• Cell Based• PDH Based (ATM cells mapped into PDH signals) (59 columns and 9 rows frame). Frame at 34.368 Mbps.• FDDI based or 100 Mbps (same as in FDDI PMD uses multimode fiber and line coding of 4B/5B). (called TAXI interface). Early private UNI interfaces were based on TAXI interfaces.• DS-3 (45 Mbps) Transfer of ATM cells on T3 (DS-3) public carrier interface. It is cheaper than SONET links. • STS-3 (155 Mbps) over Multimode fiber uses line coding of 8B/10B.• STS-3 (155 Mbps) over Twisted Pair (using Taxi interface) uses line coding of 8B/10B.• D1-T1 carriers (1.5 Mbps)

ATM INTERFACES

This interface consists of a continuous stream of cells where each cell contains 53 octets.

• Synchronization achieved through HEC basis.• Maximum spacing between successive physical layer cells is 26 ATM layer cells.• After 26 consecutive ATM layer cells, a physical layer cell (idle cells or OAM cells) is enforced to adapt transfer capability to the interface rate.

CELL BASED INTERFACE

26 0 1 26 0 1

Physical layer OAM cell

A. Transmission Frame Adaptation

•Adapts the cell flow according to the used payload structure of the transmission system in the sending direction.

•In the opposite direction, it extracts the cell flow out of the transmission frame.

Transmission Convergence Sublayer (TC)

• After initialization receiver is in the “Correction Mode”

• Single bit error detected corrected• Multiple bit error detected cell discarded

• Receiver switches to “Detection Mode”• In “Detection Mode”, each cell with a detected single-bit error is discarded.

• If a correct header is found, receiver switches to “Correction Mode”

NOTE: A noise burst of errors or other events that might cause a sequence of errors!!

B. Header Error Control (HEC)

Correction Mode Detection ModeNoError

Error detectedCell discarded

Multiple-bit errror (Cell discarded)

CorrectionSingle-bit error

No error

Example:

p Probability that a bit is in error

(1-p) Probability that a bit is NOT in error

p40 Probability that 40 bits are in error

(1-p)40 Probability that 40 bits are correct

a) With what probability a cell is rejected when the HEC state machine is in the "Correction

Mode"?

Correction Mode

Probability of a cell being rejected

Different Perspective: When is a cell accepted?

* Probability of having no errors in cell headerOR

* Probability of having a single bit error in cell header

b) With what probability a cell is rejected when the HEC state machine is in the "Detection Mode"?

Detection Mode

HEC will only accept ERROR-FREE cells.

Different Perspective:

What is the probability that a cell header is correct?

c) Assume that the HEC state machine is in the “Correction Mode.” What is the probability that n successive cells will be rejected, where n >= 1 ?

Correction Mode

Probability of n successive cells being accepted (n>1)

n=1:

Probability that 1 cell is accepted, i.e., the entire header is error-free.What is that probability?

ORThere is at most one bit error in the header.What is that probability?

n=2:

Probability that the cell header (2) is correct ANDPrevious case for cell 1

ORProbability that the cell header (2) has at most 1 bit error ANDProbability that the cell header (1) is correct (error free)

12

n=3:

Probability that the cell header (3) is correct ANDPrevious case for cell n=1

ORProbability that the cell header (3) has at most 1 bit error ANDProbability that the cell header (2) is correct ANDThe case for n=1

123

d) Assume that the HEC state machine is in the “Correction Mode.” What is the probability p(n) that n successive cells will be accepted, where n >= 1 ?

First cell is rejected:What is the probability that a cell is rejected? Case a)

Different Perspective:

Probability that all header bits of a cell are correct

Probability that one single bit error in a cell header

Remaining n-1 successive cells:

Now, HEC is in Detection Mode

What is the probability that (n-1) successive cells are rejected, i.e., there will be errors in the headers for the remaining (n-1) cells

EFFECT OF ERROR IN CELL HEADER

Incoming Cell

Error inHeader?

Valid cell(intended service)

No

Yes

Error detected

No Apparently valid cellWith errored header(unintended service)Yes

Current mode?

DetectionDiscarded Cell

CorrectionError

incorrectable?Yes

NoCorrection

attemptUnsuccessful

Successful

• Every ATM cell transmitter calculates the HEC value across the first 4 octets of the cell header and inserts the result in the fifth octet (HEC field) of the cell header.

• The HEC value is defined as “the remainder of the division (modulo 2) by the generator polynomial x8+x2+x+1 of the product x8 multiplied by the content of the header excluding the HEC field to which the fixed pattern 01010101 will be added modulo 2.”

• The receiver must subtract first the coset value of the 8 HEC bits before calculating the syndrome of the header.

• Device always preset to 0s. [Key Word: CRC (Cyclic Redundancy Check Algorithm)]

HEC Generation Algorithm (I.432)

ATM CELL STRUCTURE

1

2

3

4

5

:

:

53

8 7 6 5 4 3 2 1

8 7 6 5 4 3 2 112345:::

53

PAYLOAD(48 octets)

HEADER(5 octets)

Octet

GFCVCIVPI

VPIVCI

VCI PT PR

HEC

PAYLOAD(48 octets)

The HEC field contains the 8-bit FCS (Frame Check Sequence) obtained by dividing the first 4 octets (32 bits) of the cell header multiplied by 2^8 by the CRC code (generator polynomial)

(x8+x2+x+1)

HEC GENERATION ALGORITHM

This HEC code can 1)Correct single bit errors2)Detect multiple bit errors

REMARK: If a code corrects “t” errors, it can detect (2t + 1) errors!!!!!

i.e., Here (up to 3 bits)

•Protects the header control information•Helps to find a valid cell (cell delineation and boundaries)

Purpose:

HEC Generation Algorithm (I.432)

CELL DELINEATION

(This process allows identification of cell boundaries)

HUNT

SYNCH

PRESYNC

Correct HEC

Incorrect HEC

Cell-by-CellBit-by-Bit

consecutiveincorrect HEC

consecutivecorrect HEC

• In Hunt State a cell delineation algorithm is performed bit-by-bit to determine if the HEC coding law is observed (i.e., match between received HEC and calculated HEC).

• Once a match is achieved, it is assumed that one header has been found and the method enters the PRESYNCH state.

• The HEC algorithm is performed cell-by-cell. If consecutive correct HECs are found, SYNCH state is entered; if not the system goes back to HUNT state.

• SYNCH is only left (to HUNT) state if consecutive incorrect HECs are identified.

Cell Delineation (cont.)

Cell Delineation (cont.)

and are design parameters that influence the performance of cell delineation process.(=7 and =6).

• Greater values of result in longer delays in recognizing a misalignment but in a greater robustness against false alignment.

• Greater values of result in longer delays in establishing synchronization but in greater robustness against false delineation.

Cell Delineation (cont.)

Remarks:• A 155.520 Mbps ATM system will be in SYNCH state for more

than 5349 years even when the bit error probability is BER=10-4.

• This method may fail if the header HEC occurs in the info field (maliciously or accidentally) Cell Payload Scrambling.

• To overcome the info field contents scrambled using a self-synchronizing scrambler with polynomial X43 + 1. Header itself is not scrambled.

The probability of 7 consecutive incorrect HEC withBER=10-4

A= The probability that 7 consecutive cells are in error.

[1- (1-10-4)40 ]7 = 1.616*10-17 = A

1/A The number of cells sent in order to

have a 7 consecutive error cells; (Unit Cells);

How often does event A occur in terms of ATM cells.

{53 * 8} / {155.52 Mbps} = C

(53*8) = # of bits/cell ; Link Speed = # of bits/sec

C is how long does it take to send one ATM cell through the 155 Mbps link.

k = [1 / A] * C ={6.187*106} * {53 * 8 / 155.52 Mbps} = 1.6868*1011

k in terms of seconds

k / (365*24*60*60) approx. 5349 years..

Cell Rate Decoupling(Speed Matching)

•Adapts cell stream into Transmission Bit Rate (Insertion / Discarding idle cells; in particular for SONET Interface). SONET uses synchronous cell time slots!

Note: Cell Based Interface No need for this function.

Cell Rate Decoupling (cont.)(Speed Matching)

Buffer

+

VPI/VCI

VPI/VCI

VPI/VCI

-

Insert Idle or Unassigned cellsRemove the Idle or Unassigned cells

ATM Transmitter ATM Receiver

Transmitter multiplexes multiple streams; queueing them if anATM cell is not immediately available. If the queue is empty, when the time arrives to fill the next synchronous cell time slot, then the Transmission Convergence Sublayer inserts an Idle cell (or the ATM layer inserts an Unassigned cell.)

• Cell Multiplexing/Demultiplexing

• Cell VPI/VCI Translation

• Cell Header Generation/Extraction

• GFC Function

ATM Layer Functions

i) CELL MULTIPLEXING/DEMULTIPLEXING

In the transmit direction, cells from individual VPs

and VCs are multiplexed into one resulting stream.

At the receiving side the cell demultiplexing function

splits the arriving cell stream into the individual

cell flows appropriate to the VP or VC.

ATM Layer Functions

ATM Layer Functions

ii) CELL VPI/VCI TRANSLATION

- At ATM switching nodes, the VPI and VCI translation must be performed.

- Within VP switch, the value of the VPI field of each incoming cell is translated into a new VPI value for the outgoing cell.

- At a VC switch, the values of the VPI as well as the VCI are translated into new values.

ATM Layer Functions

iii) CELL HEADER GENERATION/EXTRACTION

- This function is applied at the termination points of the ATM layer. - Transmit Side: After receiving the cell information from the AAL, the cell header generation adds the appropriate ATM cell header except for the HEC values. HEC is done at Physical Layer. VPI/VCI values could be obtained by a translation from the SAP identifier. - Receive Side: The cell header extraction function removes the cell header. Only the cell information is passed to the AAL. - This function could also translate a VPI/VCI value into a SAP identifier.

ATM Layer Functions

iv) GFC FUNCTIONS - Supports the control of the ATM traffic

flow in a UNI. It can be used to alleviate short overload conditions.

- Control of cell flows toward the network but not flow control from the network.

- No effect within the network.

Virtual Path and Virtual Circuit Concept

•ATM cells flow along entities known as VIRTUAL CHANNELS. A VC is identified by its virtual circuit identifier (VCI).

VC set up between 2 end-users (like VC in X.25 => Indiv. Log connection).

VP Bundle of VCs having the same end points (Group logical connection; reserved trunk of connections).

•All cells in a given VC follow the same route across the network and are delivered in the order they were transmitted.

•VCs are transported within Virtual Paths (VPs). A VP is identified by its virtual path identifier (VPI). VPs are used for aggregating VCs together or for providing an unstructured data pipe.

Virtual Path and Virtual Circuit Concept

• Optical links will be capable of transporting hundreds of Mbps where VCs fill kbps. Thus, a large number of simultaneous channels have to be supported in a transmission link. Typically 10K simultaneous channels are considered (thus, VCI field up to 16bits).

• Since ATM is connection oriented, each connection is characterized by a VCI which is assigned at Call-Set-Up.

• When connection is released, VCI values on the involved links will be released or can be reused by other components.

VIRTUAL PATH / VIRTUAL CIRCUIT CONCEPT

VCI =1 (text)

VCI =2 (voice)

VCI =3 (video)

TRANSMISSION PATH

Text

Voice

Video

ATM Network Interface

Virtual Path

VP

VC

VIRTUAL PATH/VIRTUAL CIRCUIT CONCEPT

• Each VP has a different VPI value and each VC within a VP has a different value.

• Two VCs belonging to different VPs at the same interface may have identical VCI values.

• VPI is changed at points where a VP link is terminated.

• VCI is changed at points where a VC link is terminated.

Goal Multimedia Communication

Video & Voice Time Sensitive (Delay bounds) Data Loss Sensitive (Loss bounds) Allows the network to add or remove

components during the connection

e.g. Video Telephony Start with voice (only single VC)

Add video later (on another VC)

Add data (on another VC)

Signaling (on another VC)

A B

T1 T2

T3

EXAMPLE

A B

p p2

q q2

r r2

• Three VP connections exist from A to B. They are seen by A as corresponding to the values p, q, r of the VPI field, and by B as corresponding to the values p2, q2, r2. Whenever A wants to send some information to B on the VP connection seen as p, it writes the value p in the VPI field of the cell.

• The VP switches T1, T2 and T3 swap the VPI labels according to the lookup tables. The VCI field is not changed by the VP switches, so it can be used by A to multiplex several VC connections on any one of the three VP connections. Therefore, at the VC level, A has at its disposal three direct links to B.

VP Level VC Level

p

p1

p2

q q2

r r2

SWITCHING OF VCs and VPs

•Routing functions for VPs are performed at a VP switch.

•This routing involves translation of the VPI values of the incoming VP links to the VPI values of the outgoing VP links. VCI values remain unchanged.

•VC switches terminate both VC links and necessarily VP links.

•VPI and VCI translation is performed.

VP Switch/Cross Connect

VPI1

VPI2

VPI3

VPI4

VPI5

VPI6

VCI 21

VCI 22

VCI 23

VCI 24

VCI 25

VCI 24

VCI 23

VCI 24

VCI 25

VCI 24

VCI 21

VCI 22

VP Switching

VCI 23

VCI 24

VPI 2

VCI 25

VCI 21

VPI 4

VPI 5

VCI 23

VCI 24

VCI 25

VCI 21

VC Switch/Cross Connect

VP and VC SWITCHING

MORE ABOUT VCs and VPsA VP Connection:•Contains multiple VC connections.•VC connections may be built up of multiple VP connections.

•Use of VPI simplifies routing table lookup.Virtual Channel Connection

A BTD1 D2 D3 D4

Virtual Path Connection x Virtual Path Connection y

VCI = a1 VCI = a2

VPI=x1 VPI=x3VPI=x2 VPI=y3VPI=y2VPI=y1

Virtual Channel View

A T B

VCI=a1 VCI=a1 VCI=a2 VCI=a2

Other VCI Other VCI Other VCI Other VCI

VCs and VPs (Cont.)• The inter-networking of the VP and VC switches is illustrated in Figure.

• There exist VP connections (x and y) between A and T; T and B.

• Assume now that A wants to setup a VC connection to B using those two VP connections.

• The network has to provide a VCI value, say a1, for the A to T link, and a VCI value, say a2, for the T to B link.

• The VC connection from A to B is thus made of two VC links only.

• At switching points D1 through D4, only the VPI field is swapped.

• At the switching point T, both VPI and VCI fields are swapped.

• The situation is thus similar to that where A and B would be access nodes in a circuit switched network, T would be a transit node, and D1 through D4 would be cross-connects.

ATMNode 1

ATMNode 2

ATMNode 3

A B

C

VPI=4, VCI=1,2,3

VPI=6VCI=3,4

VPI=2VCI=3,4

VPI=6VCI=1,2,3

VPI=6VCI=1,2,3

VPI=8VCI=3,4

26VPIOUTVPIIN

VPIOUTVPIIN

68

46 64VPIOUTVPIIN

Example for VCIs and VPIs

• A VP is established between Subscriber A and Subscriber C transporting 2 individual connections, each with a separate VCI.

• Remark: The VCI values used (1,2,3 and 3,4 in the example) are NOT translated in the switches, which are only switching on the VPI field.

Namings

VC Virtual Channel Virtual Circuit

VC Link A point where a VCI value is assigned to another where that value is translated or terminated.

VC Identifier A value which identifies a particular VC link for a given VP Connection.

VCC (Virtual Channel Connection) A concatenation of VC links that extends between 2 points. (cell sequence integrity preserved)

VP

Bundle of VCs. VP Link

A group of VC links, identified by a common value of VPI, between a point where a VPI value is assigned and the point where that value is translated as terminated.

VP Identifier

Identifies a particular VP Link. VPC (Connection)

A concatenation of VP Links.

PVC and SVC Permanent Virtual Circuits (PVC)

Established by a network operator in which appropriate VPI/VCI values are programmed for a given source and destination (for long time).

VPs 0, …, 256 (manually configured)PVCs are established by provisioning & usually last a long time (months/years).

Switched Virtual Circuits (SVC)Established automatically through a signalling protocol (Q.2931B) and lasts for short time (minutes/hours).VCs 0, …, 65535 (automatically configured)

SOFT PVC

Part of the connection is permanent and part of it is switched.

Hybrid of PVC and SVC!!!

VCC 0 - 31 0, 5 Call set up (Signalling)

0, 16 Network Management (Integrated Local Management Interface ILMI)

32 - 65535 User Data 0, 17 For LAN Emulation Configuration

Server (LECS)

0, 18 For Private NNI (PNNI) 0, 19 or 0, 20 Reserved for future use.

Advantages of VP/VC Concept

Simplified Network Architecture: Network transport functions can be separated into those related to an individual logical connection (VC) and those related to a group of logical connections (VP).

Increased Network Performance and Reliability: The network deals with fewer, aggregated entities.

Reduced Processing and Short Connection Setup Time: Much of the work is done when the VP is set up. The addition of new VCs to an existing VP involves minimal processing.

Enhanced Network Services: The VP is used internal to the network but is also visible to the end user. Thus, the user may define closed user groups or closed networks of VC bundles.

ATM Adaptation Layer (AAL)

AAL is responsible for adaptation of information of higher layers to the ATM cells (in the transmission direction) or adaptation of ATM cells into the information of the higher layer (receiver direction).

AAL is subdivided into two sublayers:- SAR (Segmentation and Reassembly)- CS (Convergence Sublayer): Multiplexing, loss detection, timing recovery, message identification

ATM Adaptation Layer (AAL)

AAL provides a variety of services: Class 1: Circuit Emulation with Constant Bit Rates (CBR).

Voice of 64 kbps Fixed Bit Rate (Voice,Video)

Class 2: Connection-oriented service with Variable Bit Rates

(VBR) and timing between source and destination.

VBR Video & Audio

Class 3: Connection-Oriented Service.

Data Transfer and Signaling ABR Traffic with no timing

Class 4: Connectionless Data Service SMDS, Ethernet, Internet, Data Traffic,

No constraints.

Traffic ClassesA DCB

Yes No

Constant Variable

ConnectionlessConnection Oriented

1 2 ¾ or 5 ¾ or 5

TimingBetween

Source andDestination

Bit Rate

ConnectionMode

AAL

ExampleDS1, E1

N64 KbpsEmulation

Packet Video, Audio(Real Time)

FrameRelay

IP,Ethernet

General Structure of AAL

AAL-SAP

ATM-SAP

CONVERGENCE SUBLAYER(CS)

SAR SUBLAYER

PrimitivesAAL

ConvergenceSublayer

(CS)

Segmentatioin&

Reassembly(SAR)

Sublayer

Service Access Point

• Service Data Unit (SDU) crosses the SAP• PDU is data unit between peer layers

General Data Unit Naming Convention

AAL-SAP

AAL-SDU

CS-PDU PayloadCS-PDUHeader

CS-PDUTrailer

SAR-PDU PayloadSAR-PDUHeader

SAR-PDUTrailer

ATM-SAP

ATM-SDU

Cell information Field(Cell Payload)

CellHeader

PL-SAP

CS-PDUNo SAP is defined between CS and SAR

SAR-PDU

ATM CellATM Layer

Physical Layer

SegmentationAnd Reassembly(SAR) Sublayer

ConvergenceSublayer(CS)

AAL Interfaces

AAL

Segmentationof CS-PDU

Structure of AAL with SSCS and CPCS

Service SpecificConvergence Sublayer (SSCS)

Service SpecificConvergence Sublayer (SSCS)

AAL-SAP

ATM-SAP

Common PartConvergence Sublayer (CPCS)

Common PartConvergence Sublayer (CPCS)

Segmentation And Reassembly (SAR)

Segmentation And Reassembly (SAR)

AAL-PDU Primitives

SSCS-PDUPrimitives

SAR-PDU Primitives

CPCS-PDUPrimitives

SS

CS

CP

CS

C SS

AR

AA

L C

omm

on P

art

(CP

)

AA

L

AAL Type 1AAL 1 provides the foll. services to the AAL users:

• Transfer of service date unit with a constant source bit-rate and their delivery with the same bit rate

- Voice traffic 64kbps: as in N-ISDN to be transported over an ATM network. This service is called circuit emulation. In other words, how TDM type circuits can be emulated over ATM.

• CBR-Voice; CBR-Video (fixed (constant) bit rate video)

AAL Type 1

• Transfer of timing information between source and destination.

• Transfer of structure information between source and destination; some users may require to transfer of structured data, e.g., 8 kHz structured data for circuit mode device for 64 kbps (N-ISDN). • Indication of lost or errored information which is not covered by AAL1, if needed.

AAL Type 1 (Cont.)

The functions listed below may be performed in the AAL in order to enhance the layer service provided by the ATM layer:

• Segmentation and reassembly of user information

• Handling of cell delay variation to achieve constant rate delivery (playout buffer)

• Handling of cell payload assembly delay

• Handling of lost and misinserted cells (SN processing) Discarded

• Source clock frequency recovery at the receiver - 4 bit RTS is transferred by CSI - handling of timing relation for Asynchronous transfer (SRTS Synchronous Residual Time Stamp)• Monitoring of AAL-PCI (Protocol Control Information) for bit errors• Handling of AAL-PCI bit errors

AAL Type 1 (Cont.)

PCISAR-PDU HeaderCS-PDU HeaderCS-PDU Trailer

Monitoring of the user information field for bit errors and possible corrective action

- FEC maybe performed for high quality video or audio (124,128 Reed Solomon code)

AAL Type 1 (Cont.)

AAL Type 1 (cont.)

Receiver’s Responsibilities are as follows.

• Examine the CRC and parity bit for error detection.• Correct single bit errors in SN field. If multiple bit errors in SN field, then declare invalid.• Reassemble the CS-PDU in correct sequence using SN-numbers.• Discard misinserted CS-PDUs and generate dummy information for missing CS-PDU.

AAL Type 1 (Cont.)

• Buffer the received CS-PDUs to compensate for cell delay variation (jitter) to achieve constant rate delivery. (PLAYOUT Buffer)

• Clock frequency recovery (Handling of timing relationship for asynchronous circuit transport)

• Monitoring and handling AAL-PCI (Protocol Control Information) SAR-PDU Header, SAR-PDU Trailer, CS-PDU Trailer are collectively called AAL-PCI.

STACK

ATM

Convergence Sublayer

- accepts 124-byte blocks from user- appends 4-byte FEC- writes to matrix “row”- forwards CS-PDU to SAR when 47 blocks (rows) have been written

Segmentation/Re-assembly Sublayer

- reads matrix “columns”(47bytes)- effect: interleaving

AAL 1

• Forward Error Correction

• No Retransmission

* (124,128) Reed-Solomon Code* Polynomial undefined* Corrects 2 errored bytes per row* Corrects 4 “erasure” bytes (knows position)* Uses interleaving

FEC in AAL1

Cell 1Byte 1

Cell 2Byte 1

Cell 1Byte 2

Cell 1Byte 47

Cell 2Byte 2

Cell 2Byte 47

Cell 124Byte 1

Cell 124Byte 2

Cell 124Byte 47

Reading R-S Code with 4 byte FEC

Reed-Solomon Code recovers up to 4 lost cells in a block of 128.

AAL 1

User Data Bit StreamUser Data Bit Stream

CPCS-PDU PayloadCPCS-PDU Payload

AAL-SAP

…SAR-PDUPayloadH

SAR-PDUPayload

H SAR-PDUPayload

H

ATM-SAP

Cell Payload

1B 47B

48 Bytes

H Cell PayloadH Cell PayloadH…

53 Bytes

5B

Higher Layers

CS

SAR

ATMLayer

AAL

Cell Header

SN SNP SAR-PDU Payload

4 bits 4 bits

SAR-PDU Header

1 Octet 47 Octets

SAR-PDU (48 Octets)

SN (Sequence Number) for numbering of the SAR-PDUs

SNP (Sequence Number Protection) to protect the SN field

To detect lost or mis-inserted cells (Error Detection & Correction)

SAR-PDU of AAL 1

CSI Sequence Count

CRC Even Parity

13)( xxxG

13

)( xxxG

1 bit 3 bits 3 bits 1 bit

SN Field SNP Field

SAR-PDU Header of AAL 1

CSI Field

• Sequence Count 0, .., 7 • CSI bit used to transfer TIMING or DATA STRUCTURE information.

• CSI values in cells 1,3,5,7 are inserted as a 4-bit timing value.

• In even numbered cells 0,2,4,6, CSI used to support blocking of info. from a higher layer.

• If CSI bit is set to 1 in a cell 0,2,4,6, then the first octet of SAR-PDU payload is a pointer that indicates the start of the next structured block within the payload of this cell and the next cell, i.e., 2 cells (0-1, 2-3, 4-5, 6-7) are created as containing a 1-octet pointer and a 93-octet payload and pointer indicates where in that 93 octet payload is the first octet of the next block of data.

P & Non-P Formats AAL-1 CS uses a pointer to delineate the

structure boundaries. Supported by 2 types of CS_PDUs called Non-P & P

Can be used only in SAR PDUs with even SN values (because SRT scheme uses the CSI bits in SAR PDUs with odd SN values)

SN SNP SAR-PDU Payload (User Data) Non P-format

1 Octet 47 Octets

SAR-PDU Header

SAR-PDU (48 Octets)

SAR-PDU Payload SN SNP

1 Octet 1 Octet 46 Octets

SAR-PDU Header

(CSI = 1)

Reserved for Pointer

P-format

SAR-PDU

Header

Structure

Pointer Field

User

DataP-Format

Sequence Counter

0,2,4,6Reserved Bit

Offset Field

7 Bits are the offset measured in Bytes between the end of the pointer field & start of the structured block in 93 bytes consisting of remaining 46 bytes in this CS-PDU & 47 Bytes of the next CS-PDU. This offset may range from 0-92.

7 Bits

SN even uses

1 Octet Pointer field to indicate the offset into the current payload of the first octet of a n*DSO payload.

Value of n may be as large as 92 in the P-format since pointer is repeated every other cell when supporting AAL 1.

AAL1

Unstructured Data Transfer

STD Mode(Structured Data Transfer)

DS1/E1(1.544Mbps)

DS3/E3(45Mbps)

n x DSO (64kbps) Service

(supports an octet structured

n – DSO Service)

including timing SRTS Method

(4-bit RTS included in CSI Bit !!)One sent in (1,3,5,7)

CSI bit (in even SN values) for SDT to convey information about internal byte alignment structure of the user data bit stream.

Structured Data Transfer

• Kind of fractional DS1/E1 service where the user only requires an n*64kbps (DS0) connection where n can be small as 1 and as high as 24 for DS1 (T1) and 30 for E1.

• An n*64 kbps service generates blocks of n bytes which are carried in P and non-P format CS-PDUs.

• The beginning of a block is pointed to by the pointer in the 1-byte header of the CS-PDU-- > P format.

1 192 1 1 1 1 1 1 1192 192 192 192 192 192 192

193 175 18 193 165 28 193 147 46 193 137

p

47=376

SN=0CSI=1

P-Format

46=368

SN=1CSI=0

Non-P-Format

46=368

SN=2CSI=1

P-Format

47=376

SN=3CSI=0

Non-P-Format

0-1 = 93 Octets 2-3=93 Octets

Pointer indicates where in that 93 octet payload is the first octet of the next block of data.

No structured boundary, then use dummy offset value of 127.

EXAMPLE: STRUCTURED DATA TRANSFER

DS1 Signal

CS-PDUs

Unstructured Data Transfer

• The entire DS-1/E1 signal is carried over an ATM network.

• The DS-1 signal is received from user A which is packed bit-by-bit into the 47-byte non-P format CS-PDU which then becomes the payload of a SAR-PDU.

DS1 CIRCUIT EMULATION USING AAL 1

Transmitter uses AAL 1 operating in SRTS mode to emulate a DS 1 digital bit stream created by a video codec. DS1 frame has 193 bit frames that repeat 8000 times per second (192 user data bit + 1 framing bit). CS computes the RTS every 8 cell times and provides this to the SAR sublayer for insertion in the SAR header. 193 bit frames are packed into 47 octet SAR-PDUs by SAR layer. SAR then adds the SN, inserts the data from CS, computes CRC and parity over SAR header and passes 48-octet SAR-PDU to ATM layer.

octets

HeaderHeader

SAR-PDU

SAR-PDU

octets

1

47

5

48

bits1

192

1

192

192

192

192

192

1

1

1

1

Time

DS1 Signal SRTS CS SAR-PDUsATM Cells

HeaderHeader

SAR-PDU

SAR-PDU

octets

1

47

5

48

HeaderHeader

SAR-PDU

SAR-PDU

octets

1

47

5

48

RTS

EXAMPLE: UNSTRUCTURED DATA TRANSFER

Handling of Lost and Misinserted Cells in AAL1

• At the transmitter, CS provides SAR with a Sequence Count Value and a CSI associated with each SAR-PDU payload. Sequence Count Value starts with 0, and incremented sequentially and is numbered modulo 8.

• At the receiver, CS receives Sequence Count, CS indication from SAR, and check status of Sequence Count and CS indication. CS identifies SAR-PDU payload sequence SAR-PDU loss, and SAR-PDU misinsertion.

• CSI is used to transfer timing information and default value of CSI is “0”. 4 bit RTS is sent in odd-sequence-numbered PDUs (1,3,5,7) in SRTS approach.

Handling of Lost and Misinserted Cells in AAL1

Remark:

• For each SAR-PDU, SAR receives a sequence number (SN) value from CS.

• At the receiver, SAR passes the SN to CS. The CS may use these SNs to detect lost or misinserted SAR-PDU payloads.

• SAR protects the SN value and CSI against bit errors. It informs the CS when SN value and the CSI are in error and cannot be corrected.

• Transmitter computes the CRC value across the 4 bits of SAR-PDU header and inserts into CRC field. CRC contains the remainder of the division (mod 2) by polynomial of the product multiplied by the contents of SN field.

• After completing the above operations, transmitter inserts the even parity bit. 7 bit code word protected.

3x13 xx

TIMING (CLOCK) RECOVERY TECHNIQUES IN AAL 1

1. Adaptive Clocking in AAL 1

(No Network clock is available).

2. Synchronous Residual Time Stamp Approach (SRTS)

(Global Network Clock is available)

Common network reference clock is not available!!!

1. Adaptive Clocking (Receiver)

Cells PLAYOUT BUFFER

Used for Transfer

Delay Variable

Receiver reads info. with a local

clock.

Receiver writes received infofield in this buffer.

PLL (Phase Lock Loop)Provides local clock.

(Content) Filling level ofthe buffer is used to controlthe frequency of the local clock.

CONTROL is performed by continuously measuring the fill level around its median position & by using this measure to drive the PLL providing the local clock.

The content level of the buffer may be maintained within an upper limit and lower limit to present buffer overflow and underflow.

Underflow => PLL slowed down Overflow=> PLL speeded up

Network

Data

FIFO Data Terminal

Filling Level

Local Clock Jitter

Filter PLL.

Adaptive Clocking in AAL 1

Synchronous Residual Time Stamp (SRTS) Approach

BASIC IDEA: Convey a measure of the frequency difference between the reference clock and source clock. Network reference clock is available, source clock is not syncronized!

Sender NETWORKNETWORK

Common Network

Clock

Receiver

Difference between the localand network clocks.

Local Clock

TIMESTAMPCSI fieldSequence # field

Odd # ofsegments

Differencebetween 2

clocks

Transport this info. in odd numberedCells (CSI Field) to destination

LocalClock

Source Transmitter

(Assumed)• Common Network clock is available

• Source (local) clock is not synchronized with it.

• SRTS method conveys a measure of the frequency difference between the derived network reference clock and the source (local) clock.

• The derived network reference clock is determined from the frequency of the network clock divided by some integer.

• Within a time interval of N “source clock cycles” suppose there are M cycles of the derived “network reference clock”.

• There is a nominal value Mnom (fixed and known for the service) and the actual value of M may vary anywhere within a certain range (Mmin & Mmax) around this nominal value Mnom.

• The actual value of M will be the sum of Mnom and a residual part.

• By transmitting the residual part, the receiver has enough info to construct the source clock.

t

Tolerance

Source Frequency (fs)

Source clock

N cycles

T seconds

t

Derived Network

Frequency (fnx)

M M Mmin nom max

y y

24

Residual value M

Mnom

1

N

Sample&

Hold

fs

4 BitCounter

Ct

1

X

fn

fnx

Network Reference clock frequency fn is divided by x such that

1 < fnx/fs < 2

4 Bit SRTS encoded in

CSI bit for SAR-PDUs

with Sequence

Numbers 1,3,5,7

• Source clock fs is divided by N to sample the 4-bit counter Ct driven by the network clock fnx once every N = 3008 = 47 x 8 x 8 bits generated by the source.

• This sampled counter output 4 bits (residual part) is transmitted as the SRTS in SAR-PDU. • It is sent in the CSI bits of SAR-PDUs which have odd SN values.

• The method can accept a frequency tolerance for source frequency of 200 parts per million (ppm).

Ct, X, Mnom, N, fn are available at the destination and the clock value can be recovered accordingly!!!!

• For low bit rate communications, e.g., for compressed voice traffic.

• Main Idea: multiplex many users within a single ATM VCC, where each user’s information (SDT) is carried in variable length packets with a header (3 octets) identifying the user channel with control information. (kind of variable ATM cell)

• In the minicell header, the field for user identification has 8 bits limiting the number of AAL 2 users sharing a VCC to 256.

• Short and variable length payload.

• User packet multiplexing

Minicell Header 3 octets

Payload (1-64) octets SDU

AAL 2

WHY AAL 2?

AAL 1 needs not be filled with full 47 bytes. e.g., to transmit digitized voice at a rate of 1 byte every 125 sec, filling a cell with 47 bytes means collecting samples for 5.875 msec. If this delay before transmission is unacceptable, we send partially filled cells waste of bandwidth!!!

STRUCTURE OF AAL TYPE 2

AAL SAPServiceSpecific

ConvergenceSublayer(SSCS)Common

PartSublayer

(CPS)

ATM SAP

ATMLayer

PHY SAP

• Transfer of Service Data Unit with a Variable Bit Rate• Transfer of timing information between source and destination• Indication of lost or errored information which is not covered by AAL 2

AAL-SDU

SSCS-PDUTrailer

UserPacket

SSCS-PDUHeader

SSCS-PDU AAL

ATM

CPS-SDU

CPS PacketCPS-PacketHeader

CS-PacketPayload

CPS-PDUHeader CPS-Packet CPS-

PacketPAD

CPS-PDU (48 octets)

ATMHeader

ATM Cell Payload

ATM Cell

StartField

CPS-PACKET FORMAT

CPS-Packet(48 octets default64 octets optimal)

CPS-INFOCID| PPT| LI UUI HECCPS-Packet Header

(3 octets)CPS-Packet Payload

(Variable length)

CID: Channel Identifier (8 bits):Values:

•0: Not used •1: Reserved for Layer Management (AAL2 ANP packets) •2-7: Reserved•8-255: ID of SSCS entity (valid CID values to identify channels)

CPS-PACKET FORMAT (Contd)

• CID helps to multiplex multiple AAL2 users/streams (channels) onto a single VCC (ATM connection).• Each channel is identified by the CID.• A channel is bidirectional and has the same CID value.* CID field supports up to 248 individual users per VCC.

AAL2

ATM

PHY

AAL2

ATM

PHY

ATM Network

A B C D A’ B’ C’ D’

AAL2 can multiplex several data streams

AAL-SAP

CSP

SSCSSSCS

SSCS

CID=ZCID=Y

CID=X

ATM-SAP

Functional model of AAL2 (sender side)

CPS-PACKET FORMAT (Contd)

• Packet Payload Type (2 bits): serves 2 functions:

* When PPT =/ 3, the CPS packet is serving a specific application, such as carrying voice data, or carrying an ANP packet.

* When PPT=3, the CPS packet is serving an AAL network management function associated with the management of the channel identified in the CID field.

CPS-PACKET FORMAT (Contd)

* LI: Length Indicator (6 bits) * LI specifies the number of octets (minus 1) in the variable length user payload. * LI Coding: One less than CPS-Packet payload length CPS-Packet payload length = LP => LI = LP -1

* CPS-INFO: Information (variable size: (min. 1- max. 45 or 64 octets)) 45 means that exactly one CPS packet fits inside the 48 octet ATM cell payload.

CPS-PACKET FORMAT (Contd)

* UUI: User-to-User Information (5 bits): Allows the functions of an SSCS to be specific according to a purpose.

UUI serves two purposes:

• To convey specific info transparently between CPS users, SSCS entities or layer management.• To distinguish between SSCS entities and layer management users.

Codepoints: 0-27 SSCS entities 28-29 Future use 30-34 Layer management

CPS-PACKET FORMAT (Ctd)

5 bit CRC : Generator Polynomial x5+x2+1 (excluding CPS packet payload and error correction). Detectable 1 and 2 bit errors.

HEC: Header Error Control (5 bits)

CPS-PDU FORMAT

SN P CPS-Packet CPS-Packet PADOSF

CPS-PDU Payload (47 octets)

Start Field (STF) indicates the position of the first packet

CPS-PDU (48 octets)CPS-PDU Header

OSF: Offset Field (6 bits) 6 bit pointer => Position Indication of first CPS-packet (starting point of the next CPS packet header within the cell)

Values: 0-40: First CPS packet boundary (0=Next to OSF)

47-63: No CPS packet boundarySN: Sequence Number (1 bit): mod 2 (value 1 or 0)P: Parity (1 bit) : Odd parity for STFPAD: Padding (0-47 octets)

ATMHeader

Packets are streamed into successive payloads

Cell Period

Padding: All 0’sFirst

Packet

CPS Packet

ATM Cell

Pointer in OSF points to find start of a CPS packet in cell

• OSF identifies the starting point of the next CPS packet header within the cell. • If more than one CPS packet is present in a cell, then AAL2 uses the LI in the CPS packet header to compute the boundary of the next packet.

CPS Packets

ATMLayer

User 1 User 2 User 3 User 4 User 5

48 bytes 48 bytesATMHeader

ATMHeader

EXAMPLE

PURPOSE: Accommodation of low bit rate (below 64 kbps) and delay sensitive applications into ATM networks, e.g., cellular systems.

Requirements: Short Cell Assembly Time and High Efficiency.

Rt VBR Sources

CPS PDUsSAR

STF STF

16 16 16 16 16

16161616163 3 3 3 3

1 119 19 199 10 18

…

…

…

…

PACKET CAN STRADDLE CELLS!!

H Packet 1 Packet 2 H ket 3 Packet 4Pac- Packe-

ATM Cell 1 ATM Cell 2

AAL Negotiation Procedures (ANP)

• This is the function that provides the dynamic allocation of AAL2 channels on demand.

• This function is carried out by an AAL2 layer management entity at each side of an AAL 2 link.

• This layer management entity uses the services provided by AAL2 through a SAP for the purpose of transmitting and receiving ANP messages.

• These messages are carried on a dedicated AAL2 channel with CID=1, and they control the assignment, removal and status of an AAL2 channel.

• The following types of messages have been defined: Assignment request, assignment confirm, assignment denied, removal request, removal confirm, status poll, and status response.

Open Questions:

• Timing mechanisms???• Error correction schemes?

FEC but with QoS considerations!!

AAL 3/4AAL-SAP

AAL-SDUSSCS

SSCS-PDU

SSCS-PDU Payload SSCS-PDUTrailer

SSCS-PDUHeader

CPCS-SDU

CPCS-PDU Payload CPCS-PDUTrailer

CPCS-PDUHeader

CPCS

CPCS-PDU

SAR-PDU Payload SAR-PDUTrailer

SAR-PDUHeader

ATM-SAP

SAR-PDU

SAR

AAL 3/4

Non-Assured Mode(Unreliable)

Assured Mode(ARQ Protocols)

- Go_Back_N- Selective Repeat Request

Message ModeEntire AAL-PDU needed

Stream ModeSmall AAL-PDUs allowed

a) MESSAGE MODE

• AAL-SDU is passed across the AAL interface in exactly one AAL-SDU. • This service provides transport of fixed size of variable length AAL-SDUs.

• 1:1 mapping, i.e., one SSCS-PDU consists of one AAL-SDU.

• SSCS accepts a block of information from a user and creates a SSCS-PDU.

• This includes a Header & Trailer with protocol information and padding to make the PDU an integral multiple of 32 bits.

• SAR accepts the SSCS-PDU from SSCS and segments it into N 44-octet SAR-PDUs (this last segment may contain some unused portion).

AAL-SDU

H H . . .

AAL Interface

SSCS-PDU

SAR-PDUs

H

Data

SSCS-PDU Header (4 octets)

SSCS-PDU Trailer (4 octets)

Padding octets ( 0-3 octets )

SAR-PDU Header

SAR-PDU Trailer

Unused

Message Mode

H

Message mode is used for “framed data transfer”, e.g., high level protocols and applications would fit into this category, e.g., LAPD or Frame Relay would be in message mode.

• Advantage: Detects errored SSCS-PDUs and discards them.

• Disadvantage: Requires large buffer capacity.

b) Streaming Mode

• The AAL service data unit is passed across the AAL interface in one or more AAL interface data units (AAL IDUs).

• The transfer of these AAL-IDUs across the AAL interface may occur separately in time and this service provides the transport of the variable length AAL-SDUs.

• It provides transport of variable length AAL-SDU.

• The AAL-SDU may be small as 1 octet and is always delivered as 1 unit because only this unit will be recognized by the application.

AAL SDUs

AAL Interface

SSCS-PDU

SAR-PDUs

H HH

H

Streaming mode Data

SSCS-PDU

SSCS-PDU

Header(4 octets)

Trailer(4 octets)

Padding octets(0-3)

SAR-PDU Header

SAR-PDU Trailer

Unused

• Streaming mode is used for low speed continuous data with low delay requirements which may be as small as 1 octet.

• 1 block is transferred per cell.

• Data are presented to AAL in fixed size slots.

• Advantage: Transfer delay of a message is low.

A single SDU is passed to the AAL layer and transmitted in multiple SSCS-PDUs (pipelined or streamed mode).

AAL 3/4 Details

0-65535 Bytes

SAR-PDUPayload

LICRC

Cell Payload

AAL-SAP

Higher layer

CPCS

SAR

ATM Layer

CPI Btag BASize CPCS-PDU Payload0-65535 Bytes

PAD AL

ST

LengthEtag Length

SN MID SAR-PDUPayload

LICRCST SN MID …

Cell Header

ATM-SAP

H T

48 octets

TH H T

…….

53 octets

44 44

CPI: Common Part Indicator(1 Octet)

Btag: Beginning Tag (1 octet)

BA Size: Buffer Size

Allocation (2 octets)

Length: Length of CPCS-PDU

Payload (2 octets)

AL: Alignment (1 octet)

Etag: End Tag (1 octet)

PAD: Padding (0-3 octets)

ST: Segment Type (2 bits)

SN: Sequence Number (4 bits)

MID: Multiplexing

Identification (10 bits)

LI: Length Indicator (6 bits)

CRC: Cyclic Redundancy

Check Code (10 bits)

The SAR sublayer is depicted in the Figure. The SAR sublayer accepts variable length CS-PDUs from the convergence sublayer and generates SAR-PDUs with a payload of 44 octets, each containing a segment of the CS-PDU.

ST (Segment Type)The ST identifies a SAR-PDU as containing a beginning of message (BOM), a continuation of message (COM), an end of message (EOM), or a single segment message (SSM). All BOMs and COMs contain exactly 44 octets where EOM and SSM may have variable lengths.

ST ST Field

BOM

COM

EOM

SSM

10

00

01

11

Segment Type Value

AAL 3/4 Segmentation

User DataUser Data

CPCS-H CPCS-PDU Payload CPCS-TCPCS-H CPCS-PDU Payload CPCS-T

SAR-H SAR-PDU Payload SAR-T SAR-H SAR-PDU Payload SAR-T

ATM-H ATM Cell Payload ATM-H ATM Cell Payload

SAR-H SAR-PDU Payload SAR-T SAR-H SAR-PDU Payload SAR-T

SAR-H SAR-PDU Payload SAR-T SAR-H SAR-PDU Payload SAR-T

BOM

COM

EOM

ATMCell

CPCSPDU

SARPDU

SARPDU

SARPDU

SN (Sequence Number)

The SN allows the sequence of SAR-PDUs to be numbered modulo 16. SN is incremented by 1 relative to the SN of the previous SAR-PDU belonging to the same AAL connection (numbering modulo 16).These two fields enable the segments of the CS-PDU to be reassembled in the correct sequence and minimize the effect of errors on the reassembly process (counts for lost or misinserted cells, buffer overflows, and underflows & bit errors).

MID (Multiplexing Identification)* The MID is used to identify a CPCS connection on a single ATM-layer connection.

This allows for more than one CPCS connection for a single ATM-layer connection.

The SAR sublayer, therefore, provides the means for the transfer of multiple, variable-length CS-PDUs concurrently, over a single ATM layer connection between AAL entities.

Different AAL connections on a single ATM layer connection where AAL connections must have identical QoS requirements.

Multiplexing/Demultiplexing is performed on an end-to-end basis. AAL 3/4 multiplex different streams of AAL/SDUs across a single Virtual Connection.

For CO, each logical connection between AAL users is assigned a unique MID value.

Thus, the cell traffic from up to 210 different AAL connections can be multiplexed and interleaved over a single ATM connection.

For CL service, MID field can be used to communicate a unique identifier associated with each CL user and again traffic from multiple AAL users can be multiplexed.

• From a single host to forward along the same VC and be separated at the destination.

• All sessions having the same QoS MID finds which cell belongs to which session. MID desirable Carriers charge for each connection set up and for each second for an open connection.

• If a pair of hosts have several sessions open simultaneously giving each one its own VC expensive.• If 1 VC can handle the job (enough BW use)

VC1VP

VC23 sessionsMultiplexedonto VC2

AAL 3/4 Multiplexing ExampleA data communication terminal has 2 inputs with a 98-octet packets arriving simultaneously destined for a single ATM output port using the AAL 3/4 protocol.

Two parallel instances of the CPCS sublayer encapsulate the packets the packets with a header and trailer.

These are passed to 2 parallel SAR processes that request the CPCS-PDU or two different MIDs resulting in a BOM, COM, and EOM segment for each input packet.

Since all these occurs in parallel, the ATM cells are interleaved on output.

Header

PAD

Trailer

CP

CS

Pay

load

Header

PAD

Trailer

CP

CS

Pay

load

Time

Header

Trailer

SAR-SDU

Header

Trailer

SAR-SDU

Header

Trailer

SAR-SDU

Header

Trailer

Header

Trailer

Header

Trailer

SAR-SDU

SAR-SDU

SAR-SDU

Header

Payload

5

48

Header

Payload

5

48

Header

Payload

5

48

Header

Payload

5

48

Header

Payload

5

48

Header

Payload

5

48

octets

octets

98

98

2

44

2

2

44

2

2

44

2

4

octets octets

4

4

4

26

26

CPCS-PDUs SAR-PDUs ATM CellsInput Packets

2

44

2

2

44

2

2

44

2

octets

LI (Length Indicator)

The LI contains the number of octets (binary coded) from the CS-PDU which are included in the SAR-PDU payload.

Maximum value is 44. It aids in the detection of reassembly errors such as loss or gain of cells.

CRC ( Cyclic Redundancy Check )

The CRC is a 10-bit sequence used to detect bit errors across the whole SAR-PDU.

This includes the CS-PDU segment and hence the user data.

Remainder of the division (modulo 2) by the generator polynomial of the product of x10 and the content of the SAR-PDU, including the SAR-PDU header, SAR-PDU payload and LI field of SAR-PDU.

The polynomial is G(x) = x10+x9+x5+x4+x+1.

Result of CRC calculation is placed with the LSB right justified in the CRC-field (CRC-10 to detect errors).

CPI ( Common Part Indicator )

* The CPI is used to interpret subsequent fields for the CPCS functions in the Header/Trailer.

* CPI of 0 indicates that the BAsize field contains an estimate of incoming CPCS-PDU and LI exact size.

BTag ( Beginning Tag )

Sender inserts same value in BTag and ETag for a given CPCS-PDU and changes the value for each successive CPCS-PDU.

Receiver checks the values for each successive CPCS-PDU.

It also checks the value of BTag in the CPCS-Header with the value of ETag in trailer.

BTag and ETag are set to the same value to help error detection.

BASize ( Buffer Allocation Size )

The BAsize indicates the receiver the maximum buffering requirements to receive the CPCS-PDU.

BAsize is binary encoded as number of counting units. Size of counting units is identified by the CPI field.

BAsize field estimates the incoming CPCS-PDU size in bytes. Length field contains the exact size of CPCS-PDU in bytes.

In Message Mode, BAsize value is encoded equal to the CPCS-PDU payload length. In Streaming Mode, BAsize value is encoded equal to or greater than the CPCS-PDU length.

PAD

Between end of CPCS-PDU payload and 32-bit aligned CPCS-PDU trailer, there will be 0-3 unused octets for padding makes the CPCS-PDU an integral multiple of 32 bits to make end system processing more effificient.

These are used as filler octets and do not convey any information.

It may be set to zero and its value is ignored at the receiving end.

AL (Alignment ) The AL is used to achieve 32-bit alignment in the CPCS-PDU trailer. AL field complements the CPCS-PDU trailer to 32 bits. This unused octet is strictly used as a filler octet and does not convey any information, i.e., it simply makes the trailer a full of 32 bits to simplify the receiver design. AL field should be set to 0.

ETag ( End Tag )

The ETag is used to associate the CPCS-PDU trailer with the CPCS-PDU header the transmitter will insert the same value into the BTag and ETag fields.

Reassembly Process* Normally, BOM-COMs-EOM..etc.

* The first BOM causes the AAL to note the MID and SN fields, and then look for following COMs which contain the same MID and have correctly incremented the SN fields.

* Payload is extracted from each SAR-PDU to from the CPCS-PDU. * Finally, when EOM arrives, in sequence and matching MID value, then the CPCS-PDU is complete.

* Final error checking Matching ETag & BTag and ensuring the Length field matches that the received data.

Remark:

If a BOM occurs with the MID of a current CPCS-PDU being reassembled, COM & EOM SAR-PDU’s arriving with a MID value not corresponding to a current CPCS-PDU are ignored.

Those arriving with an out-of-sequence, SN field indicates an error occurred so reassembly aborted.

Example 1Suppose a sequence of SAR-PDU is transmitted through

AAL 3/4.

1. Suppose BOM SAR-PDU is lost on the way. What happens at the receiving end?

•CS-PDU will be discarded.

BOM COM COM EOMDiscard

Detect

Btag missingLength will also can do it !

2. One of the COM SAR-PDU is lost. What happens at the receiving end?

•CS-PDU will be discarded. (same as above: violation of sequence number)

•Note: SAR layer cannot detect the problem with CS. Since it has LI field (that complete data is not received), ETAG and BTAG fields.

BOM COM COM EOMCOM

Discard

SN ViolatedBuf CS SAR will not detect the problemSN will be missingLength will also detect.

3. Special Case: Suppose COM & successive EOM & BOM are lost assuming SN is matched. What happens at the receiving end?

• 2 PDUs get concatenated into the same CS-PDU

• On the CPCS layer, Btag and Etag will be different for 2 PDUs (Error occurred). Hence, everything will be discarded.

EOMBOM COM COM EOM BOM COM COM COM BOM COM COMEOM

Concatenated

Btag & Etag will be different! Discard !!

4. 16 consecutive COM SAR-PDUs are lost. What happens at the receiving end?

• When EOM SAR-PDU is received, the CS-PDU will be discarded because it is shorter than BAsize indication (Buffer Allocation size) field.

• SAR does not recognize that SAR_PDUs were lost because it uses mod 16 SN, and hence after 16 data units, the SN is repeated. However, when EOM is delivered to the CPCS, the CPCS will check the length field in the trailer of CPCS-PDU that it has assembled and will detect the assembled data is shorter than the length field. CPCS will discard it.

BOM COM COM EOMCOM

16

SN modulo 16

BA size will detect it! SAR will not compare it!

5. Multiple 16 consecutive COM SAR-PDUs are lost. What happens at the receiving end?

• Any sequence of lost COM SAR-PDU that is multiple of 16 result same as before because mod 16 SN.

Example 2

1. EOM-SAR-PDU of the first block sequence is lost.

• The partial CS-PDU of the first block will be discarded when another BOM SAR-PDU is received.

(SAR will send an ABORT signal to CPCS to terminate the Re-assembly)

So that the CPCS can release the re-assembly buffer.

2. EOM-SAR-PDU of the first block & BOM-SAR-PDU of the second block are both lost

i) Sequence numbers or

ii) E-Tag of the CPCS trailer of the second block or

iii) Length field of the second message will catch the errors.

SN of 2 subsequent messages are randomly related (AAL is free to pick any # between 0 and 15 range for initial SN of the first SAR-PDU of a message).

Suppose first message ends with a sequence like …, 6, 7 and the next message starts with 6, 7 …, if EOM (SN=6) of first message is lost BOM (SN=7) of second message is lost, sequence will appear correctly, …,6 ,7… So in this case if SN does not help in the SAR, the E-Tag will help.

If they agree, then the length field in the CPCS-PDU of second message will catch it.

AAL 5• The new AAL was introduced in the study process of CCITT at the end of 1991.

• Its description was published in the 1994 CCITT recommendations.

• Designed for the same class of service as AAL 3/4, it has the advantage of being simpler and requiring less overhead.

• Unlike AAL 3/4, it allows all 48 octets of the cell information field to be used for the transport of CS-PDU segments, the only SAR protocol information being provided by a bit in the ATM cell header, as explained below.

AAL 5•This means that there is neither multiplexing nor error control at the SAR sublayer.

•However, there is a CRC field (CRC-32) in the CS sublayer.

•There are also similarities with AAL 3/4.

•The two modes of service defined,

message and streaming mode

are the same as in AAL3/4.

The Convergence Sublayer of AAL 5 has been subdivided into a CPCS part and a SSCS part.

• CPCS:

--supports streaming mode and message mode

• SSCS: uses the same SSCS as AAL 3/4 and provides

assured or non-assured data delivery.

• The protocol control information field of the SAR sublayer uses the ATM-layer-user-to-ATM-layer-user parameter (AAU) contained in the ATM header to indicate that a SAR-PDU contains the end of a CS-PDU.

* When the bit is set to 1, it indicates the end of the CPCS-PDU; when the bit is set to 0 it indicates the continuation or the beginning of a CS-PDU.

* This is necessary to enable the SAR to copy with reassembly of the CS-PDU in the presence of errors.

* If no indication of the end of the CS-PDU was provided, the loss of a cell, and hence the loss of a segment of the CS-PDU, would mean that all subsequent reassembly operation would be incorrect.

* By indicating the end of the CS_PDU, the loss of a single cell would limit the error to one CS-PDU, unless the lost cell contained the end indication in which case the error would be restricted to 2 CS-PDUs.

ATM CELL STRUCTURE

• Octets are sent in increasing order 1,2,3 …• Within an octet the bits are sent in decreasing order 8,7,6,5,4 ...

1

2

3

4

5

:

:

53

8 7 6 5 4 3 2 1

1

2

3

4

5

:

:

53

8 7 6 5 4 3 2 1

8 7 6 5 4 3 2 112345:::

53

PAYLOAD(48 octets)

HEADER(5 octets)

Octet

User Network Interface (UNI)Cell Structure

Network Network Interface (NNI)Cell Structure

GFC

VCI

VPI VPI

VPIVPI VCI

VCI

VCI

VCI

VCI PT PT PRPR

HEC HEC

PAYLOAD(48 octets)

PAYLOAD(48 octets)

GFC : Generic Flow ControlVPI : Virtual Path IdentifierVCI : Virtual Channel IdentifierPT : Payload TypePR : PriorityHEC : Header Error Control

PAYLOAD TYPE (PT)

First Bit 0 User Information

First Bit 1 Network Management or Maintenance Function

Second Bit Whether CONGESTION has been experienced or not.

Third Bit known as AAU (ATM-User-to-ATM-User) used in AAL5 to convey information between end users.

Remark:

• Lack of LI field No way for SAR to distinguish between CPCS-PDU octets and filler in the lost SAR-PDU. There exists no way for SAR entity to find the CPCS-PDU trailer in the last SAR-PDU.

•To avoid these situations - CPCS-PDU payload be added out so that the last bit of the CPCS trailer occurs as the last bit of the final SAR-PDU.

•No Sequence Number Receiver must assure that all SAR-PDUs arrive in proper order for reassembly. CRC should guarantee that.

• Lack of MID:

It is not possible to interleave cells from different CPCS-PDUs. (Each successive SAR-PDU carries a portion of the current CPCS-PDU or the first block of the next CPCS-PDU).

• 32-bit CRC for AAL 5

G(x)=x32+x26+x23+x22+x16+x12+x11+x10+x8+x7+x5+x4+x2

+x+1

Probability of undetected cell misordering is 2-32.

AAL 5

PAD: Padding (0 to 47 octets); Length: Length of CPCS-SDU (2 octets)CPCS-UU: CPCS user-to-user indication (1 octet)CRC: Cycle Redundancy Check (4 octets)CPI: Common Part Indicator (1 octet)

PADCPCS-PDU

Trailer

CPCS-PDU Payload

CPCS-UU CPI Length CRC

CPCS-PDU

8 Octets

SAR PDU Format for AAL5

PT (Payload Type): The PT belongs to the ATM header and it conveys the value of the ATM-layer-user-to-ATM-layer-user indication.

Cell Header

PT

SAR-PDU (48 Octets)

SAR-PDU Payload

AAL 5 Example

* Two 98-byte packets arrive simultaneously.

* Two parallel instances of this CPCS sublayer.

* Add a trailer to each packet.

* Note that the entire packet does not have to be received

before it can begin the SAR function as in AAL 3/4 to

insert the correct buffer allocation size.

* The packets are segmented by 2 parallel SAR processes.

* Here these cells are destined for the same VPI/VCI and

hence only one can be sent at a time

Input Packets

CPCS-PDUs

SAR-PDUs

ATM Cells

CP

CS

Pay

load

48

5

Time

octets

octets

octets

PAD

SAR-SDU

SAR-SDU

SAR-SDU

HeaderPayload

SAR-SDU

SAR-SDU

SAR-SDU

48

48

48

48

48

Trailer

PAD

Trailer

CP

CS

Pay

load

388

38

8

98

98

octets

HeaderPayload

HeaderPayload

HeaderPayload

HeaderPayload

HeaderPayload

48

5

48

5

48

5

48

5

48

5

48

Example1. Single bit error in 1 of the SAR-PDUs occurs.

CS-PDU will be discarded when lost SAR-PDU is received due to CRC failing. In trailer (CRC - checking CS layer AAU=1 after getting EOM-SAR

2. Suppose one of the cells with AAU=0 is lost.

Find SAR-PDU (AAU=1) will cause the CPCS check the LENGTH FIELD of the CPCS-PDU trailer. Trailer is always in the last cell (AAU=1), and CRC will ALSO be checked.

3. One of the cells AAU=1 is lost?

Error either by CRC or by mismatch of the length field in CPCS-PDU trailer OF THE NEXT ARRIVING CELL with AAU=1.

When trailer for next message (the one that is lost with AAU=1 cell) is received. This will result in a loss of both corrupted and the next message.

REMARK:

The LEN field of CS-PDU is limited to 2 bytes. So a max. of 64K bytes can be sent before the end of message error can be triggered off. If the total size of both CS-PDUs is greater than 64K, receiver will detect the error.

Sources of Cell Losses

a) Errors on the transmission media

b) Discarding cells for congestion control

c) Processing errors in switching nodes and end-points

Effect of Cell Loss on Reassembly

Cell Loss

AAL 3/4 may pass partially reassembled CPCS-PDUs to the user along

with an error indicator.

AAL 5 Can only pass up an error indicator

AAL 3/4 Reassembly

• Receiver SAR/CPCS rejects all COM & EOM cells passed to it. BOM is required. If BOM lost, the entire CPCS-PDU is discarded.

• Incorrect SN progression between SAR-PDU reveals the loss of a COM.

• If a multiple of 16 consecutive cells is lost, then the SN wraps around, but the loss of data is detected by the CPCS-PDU being undersized.

• To detect EOM loss, two methods exist:

METHOD 1:

If the BOM of the next CPCS-PDU on the same MID arrives before the EOM for the current CPCS-PDU, then the partially reassembled CPCS-PDU must be released by the SAR/CPCS.

Entire partially reassembled CPCS-PDU received to that point is considered valid & passed to the AAL user along with an error indication.

This is only the case when EOM is lost or where a cell burst knocks out some COMs followed by the EOM.

A cell loss burst that knocks out the EOM & the following BOM & slips past the SN checks, will be detected when B-tag & E-tag fields fail to match.

REMARK:

The length indicator may fail to pick up this error, if the cell burst loses as many cells as are added by concatenating the 2 CPCS-PDU fragments.

In this case only the first 44 bytes of the first CPCS-PDU may be legitimately retrieved.

For the second one ---> bad luck ....

METHOD 2:

Attach a timer to each CPCS-PDU under reconstruction & signal an error

when it is not reassembled within a certain time frame.

AAL 5

Encapsulation & Seq-Checking DO NOT EXIST as in AAL 3/4.

Reassembly Errors

are detected only when CPCS-PDU trailer arrives. Impossible to know how much has been received already is correct. Single-bit errors in SAR-PDU are not picked up until the CPCS-PDU CRC is calculated ---> if

incorrect the entire CPCS-PDU is discarded.

Lost of cells with AAU=0, detected by an incorrect CRC when the trailer arrives. If CRC fails to flag the error, the length field mismatch ensures the CPCS-PDU is discarded.

Loss of Cells with AAU=1 detected in 3 ways

• SAR-PDU of the following CPCS-PDU may be appended to the first, resulting in a CRC error (or length mismatch).

• AAL may enforce second CPCS-PDU can flag an error and cause the assembled data to be discarded.

• A timer attached to CPCS-PDU reassembly. If it expires, assembled is discarded.