M.

Vuskovic

ATM Networking

CS578

Chapter 2 ATM SIGNALING

Table of contents:

2.1 OVERVIEW 2.2 UNI SIGNALING 2.1.1 ATM End Station Addressing (AESA) 2.1.2 Automatic Addressing Registration 2.1.3 ILMI 2.1.4 Connection Setup 2.1.5 Timers 2.1.6 Connection Teardown 2.1.7 Status Inquiry 2.1.8 Point-to-Multipoint Connections 2.1.9 Add Party and Drop Party 2.1.10 Leaf Initiated Join (LIJ) 2.1.11 Signaling Messages 2.1.12 Protocol Stack for UNI Signaling 2.1.13 Protocol Planes 2.3 NNI SIGNALING 2.2.1 IISP 2.2.2 PNNI 2.2.3 Hierarchical Addressing 2.2.4 Peer Groups 2.2.5 Higher-Level Peer Groups 2.2.6 Topology Aggregation 2.2.7 Uplinks 2.2.8 Hierarchical View of a Node 2.2.9 Hello Protocol 2.2.10 Flooding Topology State Information 2.2.11 Call Setup and Routing

Copyright C 1998, 1999 by Marko Vuskovic

2-2

OVERVIEW
As shown in the previous chapter, the flow of cells between the end stations via one or several ATM switches doesn't require any routing decision in any of the nodes along the virtual connection (PVC, SVC or SPVC). The cells are switched according the cell labels (VPI,VCI) and the cell flow enjoys the agreed QoS. How is a virtual connection established and how is QoS negotiated for that connection? This is done through the process called signaling. Signaling performs call setup and call clearing, and many other functions related to the network. There are two sets of standards for ATM signaling: UNI signaling, and NNI signaling UNI signaling is performed between an end station and a private ATM switch, or between a private ATM switch and the public ATM network. The UNI signaling is simpler because it doesn't involve routing. The standards are produced by the ATM Forum and are called UNI 3.1 (1994) and UNI 4.0 (1996). (UNI 4.0 is an addition to UNI 3.1, UNI 3.1 is derived from the Public Network Signaling protocol Q.2931 brought by the ITU-T, which is further derived from Q931 used in ISDN and Frame Relay). NNI signaling is performed between the switches of a public ATM network. Since a public network generally involves several (or many) switches the routing becomes very important component of the NNI signaling. NNI signaling has two major standards: IISP (Integrated Interswitch Signaling Protocol) and PNNI (Private Network-to-Network Interface). IISP is a simple signaling protocol which uses static routing and is designed for small private ATM networks. PNNI is a signaling protocol that uses very elaborate dynamic routing algorithms which can easily handle small to large ATM networks which can have hundreds, thousands and even tens of thousands of ATM switches. This chapter discusses the basic principles and mechanisms for UNI and NNI signaling.

UNI
End Station ATM Switch

NNI
SIGNALING ROUTING
ATM Switch

UNI
End Station

SIGNALING

SIGNALING

UNI 3.1 UNI 4.0

IISP PNNI

UNI 3.1 UNI 4.0

Copyright C 1998, 1999 by Marko Vuskovic

2-3

UNI SIGNALING

ATM End Station Addressing (AESA) All ATM switches and end stations in an ATM network must have a unique ATM address. The address is a crucial part of ATM signaling. This address must be long enough to accommodate a potentially huge number of ATM devices.

20 octets (40 HEX digits)

13 octets

6 octets

1 octet

Prefix

ESI
(End System Identifierr)

SEL
(Selector)

Identifies switch to which the end system is attached

Identifies the end system

For vendor use (not for ATM, for example can identify different LAN emulation server, or can be used for local multiplexing)

ESI must be unique within the prefix. It is recommended that ESI is globally unique. Therefore the MAC address can be used

Example:

Prefix

ESI

SEL

39.246F.00.0E7C9C.0312.0001.0001.000012345678.00
Copyright C 1998, 1999 by Marko Vuskovic

2-4

UNI SIGNALING (Cont.)

There are three different AESA formats. The formats vary by the prefix alone. Authority Format Identifier - identifies authority that is allocating the Data Country Code, the International Code Designator, or the E.164 number. This field is specified by the authority identified by first two fields. It describes the hierarchy of the addressing authority and the topological significance, which is used in the ATM routing algorithms (see section "NNI signaling")

1 octet

2 octets

10 octets

AFI 39

DCC
(Data Country Code)

HO-DSP
(High-Order Domain Specific Part)

DCC AESA Format

Country codes (in BCD) as specified by ISO 3166 (For example US is coded 840). Field left justified and padded by "F"

1 octet

2 octets

10 octets

AFI 47

ICD
(International Code Designator)

HO-DSP
(High-Order Domain Specific Part)

ICD AESA Format

Similar to DCC, only it is maintained by British Standards Institute (BSI).

1 octet

8 octets

4 octets

AFI 45

E.164

HO-DSP
(High-Order Domain Specific Part)

E.164 Based AESA Format

This addressing format combines the E.164 addressing scheme (used in PSTN and N-ISDN) with the AESA. The E.164 number is padded with leading zeros and by a trailing "F". For example +01(619)594-7898 would be represented by: 000016195947898F

Copyright C 1999 by Marko Vuskovic

2-5

UNI SIGNALING (Cont.)

Automatic Address Registration AESA has 20 octets, or 40 nibbles. With such long address it would be very difficult to configure an edge device each time a device is added to the ATM network. Such a job is even difficult with IP addresses which have only four octets. In order to relieve the problem of network configuration IP has introduced Dynamic Host Configuration Protocol (DHCP). ATM uses a similar idea. The ATM addresses (prefix only) of switches must be entered manually by the network manager. Once the address is in place, each work station (edge device) attached to that switch can now be configured automatically. The configuration is dynamic, it happens each time a device is attached to the switch, or when the device is moved from one switch to another. As shown latter, a work station can have, in addition to the ATM address an IP address, which is necessary if it is running a TCP/IP application. An important characteristic of the ATM automatic addressing is that the IP address of edge devices doesn't change at its relocation, only the ATM address changes. In other words, the TCP/IP user is not aware of the relocation of the work station.

ILMI Automatic address registration is performed through the Integrated Local Management Interface (ILMI) (when ILMI was initially created by the ATM forum in 1992 it was ment as a temporary protocol which was supposed to be replaced later. Therefore, the "I" in ILMI originally was "Interim". However, the protocol has proven to be convenient and it was decided later to keep it forever, thus the initial letter now stands for "Integrated"). ILMI is based on IP's SNMP (Simple Network Management Protocol) and uses a similar Management Information Base (MIB) and access procedures like Get, Set and Trap requests and responses. All ILMI communications go over a dedicated (default) VC (VPI = 0, VCI = 16) Each ATM device (edge device or switch) that implements UNI (private or public) has ILMI and a component called Interface Management Entity (IME). This entity acts as a symmetric component that can both send requests and respond to a peer IME. IME is responsible to maintain MIB and interpret/respond to SNMP messages. There are four types of SNMP messages used in automatic address registration: Trap Get GetNext Set - Used to report extraordinary events - Used to retrieve a specigied management information - Used to retrieve a man. info. without knowing its specifier (by table search) - Used to alter the specified management information

Copyright C 1999 by Marko Vuskovic

2-6

UNI SIGNALING (Cont.)

Edge device

ATM Switch SNMP over ALL5 VPI = 0, VCI = 16

Edge device

ATM Switch Private or public UNI

IME (user)

IME (net)

Cold Start trap

Cold Start trap

"I am ready to start automatic address registration, please clear all any previously exchanged address information and invoke ILMI procedures"

GetNext request
"Did you clear your address table? If you answer 'YES' I'll proceed, if not I'll send another ColdStart trap"

GetNext request
"Here is your 13 octet prefix, please write it into your address table." "Here is my 20 0ctet address (the prefix I have gotten from you and my ESI and SEL"

Set request

Get response

NOTICE: This information echange is symetric, both UNIs (user or network) can start the exchange
Copyright C 1999 by Marko Vuskovic

2-7

UNI SIGNALING (Cont.)

Once an AESA address is established the user can place a call across an ATM network. The calls are acomplished by a set of signaling frames. There are three types of signaling frames: connection setup frames maintenance frames connection teardown frames All these frames use dedicated VC, VPI = 0, VCI = 5. The detailed structure of the signaling frames is discussed later (see subsection: "Signaling Messages").

Connection Setup

User A

UNI ATM Network

User B UNI

Call Reference Source Address Destination Address Traffic Charact. QoS

Setup Call Proceeding

Setup

- Call Reference - VPI/VCI (which will be used for data transfer, if B accepts the call)

NNI Signaling

Call Reference Source Address Destination Address Traffic Charact. QoS VPI/VCI (for data transfer

Call Proceeding Connect Connect Ack

- Call Reference

Connect
- Call Reference

Connect Ack

At this point the entire VC is established

Call Reference is a unique number that identifies the call at the local UNI (has local significance). It is assigned by the user A, and by the engress switch of the user B.
Copyright C 1999 by Marko Vuskovic

2-8

UNI SIGNALING (Cont.)

Timers Both users and network nodes use timers to define reasonable wait periods for completion of certain actions (setup, call proceeding, connect, connect ack, etc.) The ATM UNI signalling provides 10 timers at the user side and 10 timers at the network side. Examples: Timer T303 Starts when Setup message is sent; stops when Call Proceeding or Connect is received. If T303 expires before Call Proceeding is received, a Setup message may be retransmitted. (If the UNI doesn't support the Setup retransmission, the intended connection is cleared and the null state is entered.) Timer T310 Starts when Call Proceeding is received, and is stopped when a Connection or Release (see next page) message is received. If the timer expires before receiving Connection or Release, the intended connection is cleared.

User A

UNI ATM Network

User B UNI

Start T303

Setup
Call Proceeding

Start T303
Setup
Call Proceeding

Stop T303 Start T310

Stop T303 Start T310 Stop T310

Connect Connect

Start T313

Stop T310

Connect Ack

Connect Ack

NNI Signaling

Stop T313

Based on U. Black: "ATM Vol I, Foundation of Broadband Networks", Prentice Hall, 1999
Copyright C 1999 by Marko Vuskovic

2-9

UNI SIGNALING (Cont.)

Connection Teardown
User B ATM Network

User A

Start T308
Release
- Call Reference - Cause

Start T308
Release Complete

Stop T308

NNI Signaling
Stop T308

Release
Release Complete

- Call Reference - Cause

Causes for connection teardown: natural end of call, a timer has expired, one of the links along the communication path has experienced an unrecoverable error. After a natural end of call, either user can initiate a connection teardown. If the release timer expires before Release Complete is received, the Release message is retransmitted. If the timer expires again, the connection is cleared and the system enter a null state.

Status Inquiry Can be invoked by either the user or the network, in order to determine the state of the connection (call state, connection type). Operates only on the link basis.

User A ATM Network

User B

Status Inquiry

Status

Copyright C 1999 by Marko Vuskovic

2-10

UNI SIGNALING (Cont.)

Point-to-Multipoint Connections The signaling considered so far was for point-to-point connections where only one calling party and one called party were involved. Point-to-multipoint connections have one calling party (root) and many called parties (leaves). The best example of this type of connection is broadcast video. Due to the complexity, some vendors do not support point-tomultipoint connections, or limit number of leaves per connection.

Single copy of messages

Switches make copies of cells and deliver them to different leaves

Root Device (e.g. video server)

Switch

Leaf

Leaf Leaf

Copyright C 1999 by Marko Vuskovic

2-11

UNI SIGNALING (Cont.)

Add Party and Drop Party Point-to-multipoint connections are achieved through the Add Party message. First, a point-to-point connection must be established (through Setup message). Then, the root establishes other connections, one by one, using the Add Party message. One or several called parties can be dismissed from the connection by the Drop Party message. This message can be activated from both sides.

Root
User A UNI ATM Network UNI

Leaf
User B

Start T399

Add Party

Start T303
Setup
Call Proceeding

Stop T303 Start T310 Stop T310 Stop T399

Add Party Ack Connect Ack Connect

Start T313

NNI Signaling
Start T398
Drop Party Drop Party Ack

Stop T313

Start T308
Release Release Complete

Stop T398

Stop T308

Based on U. Black: "ATM Vol I, Foundation of Broadband Networks", Prentice Hall, 1999
Copyright C 1999 by Marko Vuskovic

2-12

UNI SIGNALING (Cont.)

Leaf Initiated Join (LIJ) The Add Party message can be initiated only from the root device. However, the new version of the UNI signaling protocol, UNI 4.0, provides the possibility where a leaf can initiate a connection and join an existing point-to-multipoint connection. This is accomplished with two messages: Leaf Setup Request and Leaf Setup Failure. The latter message is sent by a network node or by the root in case either refuses to accept the new connection initiated by the leaf.

Root
User A UNI ATM Network UNI

Leaf
User B

Leaf Setup Request

"I am leaf B and want to join the connection initiated by the root A"

If the root accepts the request it will issue a setup request

Leaf Setup Request

Setup

This request is forwarded to the root if there is an active connection

Call Proceeding

NNI Signaling

Setup Connect Connect Ack

Connect Connect Ack

Copyright C 1999 by Marko Vuskovic

2-13

UNI SIGNALING (Cont.)

Signaling Messages All signaling frames discussed in previous subsections have the following general format:

Octet 1
2

0 0
Flag

0 0

0 0

1 0

Protocol Discriminator (identifies Q.2931 messages) Unique number assigned by the call originator (end station, and network) Stays unchanged during the call. (Flag = 0 - originating side; 1 - destination side) SETUP, CALL PROCEEDING, CONNECT, CONNECT ACK, RELEASE, ADD PARTY...

Length of CR

3-5

Call Reference

6-7

Message Type

8-9

Message Length

10-n

Information Elements

Depends on message type

The most complex message is the SETUP message which uses the biggest number of information elements. Here are some important information elements used by the SETUP message: AAL parameters (Specify the AAL type and the relevant parameters) ATM user traffic descriptor (Forward/backward PCR, SCR, MBS) ........................................... Called party number (Usually an E.164 number) Called party subadress Calling party number (Usually an E.164 number) Calling party subadress Connection identifier (VPI, VCI) QoS parameter (To requested and indicate QoS class for connection) ...........................................
Copyright C 1999 by Marko Vuskovic

2-14

UNI SIGNALING (Cont.)

Each information element has a uniform structure. Here is an example of the QoS parameter information element:

Octet 1 2 3-4

0 1

1
CS

1 0

1 0

1 0

0 0

0 0

Information element identifier (QoS Parameter)

Length of QoS parameter contents Forward QoS Class Backward QoS Class

Coding Standard, CS = 00 - ITU-T; CS = 11 - standard defined for the network side of UNI
See definition below (valid for CS = 11, for CS = 00 reserved for ITU-T)

5 6

Individual QoS parameters can be added if QoS Class field is 'FFFF"

ITU-T Recommendation for QoS Classes (I.356) QoS Parameter CTD CDV CLR (0+1) CLR (0) CER CMR SECBR QoS Class 1 (CBR video, CE) 400 ms 3 ms 3x10-7 N/A 4x10-6 Once per day 10-4 QoS Class 2 (VBR video/audio) Unspecified Unspecified 10-5 Unspecified 4x10-6 Once per day 10-4 QoS Class 3 (CONS data) Unspecified Unspecified Unspecified 10-5 4x10-6 Once per day 10-4

CTD - Cell Transfer Delay CDV - Cell Delay Variation CLR - Cell Loss Ratio CLR(0) - guaranteed performance on CLP=0 cells CLR(0+1) -guaranteed performance for CLP = 0+1 cell stream CER - Cell Error Rate CMR - Cell Missinsertion Rate SECBR - Severely Erroded Cell Block Ratio There are also Class 0 (Unspecified QoS, no QoS guaranteed), and QoS class 4 for connection-less data transfer (IP and SMDS)
Copyright C 1999 by Marko Vuskovic

2-15

UNI SIGNALING (Cont.)

If the SETUP message is accepted (by the network at the originating side, or by the end user at the destination side), the CALL PROCEEDING or CONNECT messages are returned. However, if the call is not accepted, then the RELEASE COMPLETE message is returned and the connection is aborted after which the system enters the null state, ready to establish a new connection. The RELEASE COMPLETE message has one information element called "cause"
Information element identifier (Cause)

Octet 1 2 3-4

0 1

0
CS

0 0

1 0

0 0

0 0

0 0

Length of cause information contents

Coding Standard, CS = 00 - ITU-T; CS = 11 - standard defined for the network side of UNI
(user, private net. serving the local user, public net. serving the local user, transit net. private net. serving the remote user, public net. serving the remote user, international net., network beyond internetworking point)

5 6

1 1

Location

Cause Value Diagnostics (if any)

Examples of cause value: - unassigned number - no route to specified transit network - no route to destination - normal call clearing - user busy - no user responding - call rejected (diagnostic extension gives rejection reason) - number changed - incomplete address - response to STATUS INQUIRY - normal unspecified - requested VPI/VCI not available - network out of order - QoS unavailable - AAL parameters cannot be supported - invalid call reference value - too many panding ADD PARTY reqests - non-existing information element - message not compatible with the call state - incorrect message length - unspecified protocol error ...........................................
Copyright C 1999 by Marko Vuskovic

2-16

UNI SIGNALING (Cont.)

Protocol Stack for UNI Signaling The signaling considered in previous subsections can be understood as an application which communicates with a similar peer application which resides on the other end of UNI. Such an application is implemented by a signaling protocol called Q.2931 (specified by ITU-T), or by the equivalent called UNI 3.1/4.0 (specified by the ATM forum). As shown, this protocol deals with messages with lengths that can vary in size from 10 to over 240 octets. In order to map these messages into ATM cells, they have to go through adaptation layer in a similar way as the data packets go through AAL once the connection is established. Since the requirement for reliability is stricter than in case of data transmission (like in Frame Relay), a complete different protocol stack is used for signaling.

Q.2931 (UNI 3.1, UNI 4.0)

SSCS CS SSCF SSCOP CPCS SAR

UNI signaling (SETUP, CALL PROCEEDING, CONNECT, CONNECT ACK, RELEASE,...) Service Specific Coordination Function (Interface between Q.2931 and SSCOP) Service Specific Connection Oriented Part (reliable transport of signaling messages, error recovery by retransmission) Similar to AAL 5 for data transmission ATM layer is common for both signaling and data transmission (Uses VPI=0, VCI=5) (SAAL = Signaling AAL)

SAAL

ATM
Physical Layer

Copyright C 1999 by Marko Vuskovic

2-17

UNI SIGNALING (Cont.)

Protocol Planes As seen, there are three different activities regarding ATM networking: network management, signaling, and data transmission. So far, we have considered only a part of network management (automatic address registration) and only part of signaling (UNI signaling) - they will be discussed more in the following sections and chapters. These three actvities have different requirements and therefore have different adaptation layers. Consequently the ATM adaptation layers are divided into three different protocol stacks called planes.

Control Plane

User Plane

Management Plane

Q.2931 (UNI 3.1, UNI 4.0)

SSCS CS SSCF SSCOP CPCS SAR

TCP/IP, CLIP, LANE, CE, video/audio, etc.

SSCS CS

ILMI, SNMP, CMIP

SSCS

SAAL

AAL

CPCS SAR

AAL

CS CPCS SAR

ATM SONET/SDH, T3, E3,...

SETUP, CALL PROCEEDING, CONNECT, RELEASE, ADD PARTY

Data, voice, video

TRAP, GET, GET NEXT, SET

Q.2931 SAAL

User Appl. AAL ATM Phy

ILMI AAL Cells

ILMI AAL

User Appl. AAL ATM Phy

Q.2931 SAAL

Copyright C 1999 by Marko Vuskovic

2-18

NNI SIGNALING
Network-to-network signaling consists of protocols that implement signaling between ATM switches. An important issue in NNI signaling is routing, which takes place before the VC is established. Once the VC is established and the VPIs, VCIs are determined for each node and edge device, then the routing isn't used any longer. There are two NNI signaling protocols currently used: Interim Interswitch Signaling Protocol (IISP) Private Network to Network Interface (PNNI)

IISP is a simple protocol with static routing tables which have to be manually created and maintained. This protocol is difficult to manage and administer and it limits network scalability. IISP is used for small private ATM networks with a small number of switches. The protocol is employed mainly to fill the gap before PNNI was available. PNNI is a signaling and a dynamic routing protocol that enables scalable networks with a large number of dynamic multi-vendor ATM switches.

IISP IISP is an extension of UNI 3.1/4.0 (approved in 1994) which includes simple hop-tohop routing based on AESA addresses. Usually, the table has two additional fields for output ports: the second and the third routing choice in case the link for the first choice fails. For routing are used only the first n octets of the address (n is the column indicated by "Octets to use"). An IISP routing table must be configured by the network administrator. Octets to use Port (choice) 1st 2nd 3rd

Address (AESA - prefix)

47 0006 40FFC1000000F33B4C12 45 000016195937898F 0F12AA654 47 0006 40FFC1000000FF88912C 47 0006 40FFC1000000F78B2CF1 39 0006 40FFC1000000F33B22EE

9 1 12 12 9

1 3 2 3

2 2 3 2

3 1 1 1

Copyright C 1999 by Marko Vuskovic

2-19

NNI SIGNALING (Cont.)

A potential problem with IISP routing tables are routing loops (see example below). Dynamic routing with PNNI doesn't have these problems.

Address

Destination node

A 1 2 3
Link is down or congested

Route choice 1st 2nd 3rd ................ A 1 3 2 ................

1 2 3

1 E 2

1 2 3 1 D C
Address Route choice 1st 2nd 3rd ................ A 2 3 1 ................

Address

Route choice 1st 2dn 3rd ................ A 1 2 ................

2 3

Address

Route choice 1st 2nd 3rd ................ A 2 3 1 ................

The endless looping could be avoided if D could change the route choice after the first loop and choose the 2nd instead of the 1st choice. This however is not possible in static routing.

Copyright C 1999 by Marko Vuskovic

2-20

NNI SIGNALING (Cont.)

IISP is essentially the same as UNI signaling with some minor additions. The signaling can be done only after the path between the source and the end station has been established. In the case of IISP, the path is always known in advance due to the static nature of routing table, which is configured manually.

ATM Network
User A UNI IISP IISP UNI User B

Setup Setup Call Proceeding Setup Call Proceeding Setup Call Proceeding Call Proceeding Connect Connect Connect Connect Connect Ack Connect Ack Connect Ack Connect Ack

Copyright C 1999 by Marko Vuskovic

2-21

NNI SIGNALING (Cont.)

PNNI PNNI has two major parts: PNNI routing, and PNNI signaling PNNI routing is used to dynamically gather the information about the status and the topology of the network, then to determine an optimal route for each particular connection establishment. PNNI signaling is similar to UNI signaling, it has similar messages and initiates at the calling end station. PNNI is extremely complex (the ATM Forum's specification of PNNI is over 350 pages, just the basic document, without many addenda) The complexity is a price for many advantages that PNNI offers over the classical routing protocols used in IP. Some of these advantages are: high scalability (adding tens of thousands of nodes makes no difference) support for QoS-based routing connection-oriented communication with guaranteed bandwidth quick response to traffic conditions adaptation to changing network topology high automatism (makes the network administration an easy job)

These advantages have made PNNI the most powerfull routing protocol defined to date. The core principle used to manage the complexity and to support scalability is hierarchical representation of the network topology and hierarchical routing. The following pages will discuss the hierarchical addressing and hierarchical peer groups which are employed in PNNI.

Copyright C 1999 by Marko Vuskovic

2-22

NNI SIGNALING (Cont.)

Hierarchical Addressing The basic AESA format is shown at the beginning of this chapter. The high-order domain specific part (HO-DSP) can be used to reflect the network hierarchy (address is therefore topologically significant). Suppose that there is an ATM network that spans several cities, each city has several buildings, each building has several floors, and each floor has several machines.
Building 000001 Floor 04 Building 000002 New York (CC = 0004)

San Francisco (CC = 0001)

Cleveland (CC = 0005)

Los Angeles (CC = 0002) Houston (CC = 0003)

Then the HO-DSP part of the address prefix can be as follows:

1 octet 2 octets 10 octets

Prefix

AFI 39

DCC
(Data Country Code)

HO-DSP
(High-Order Domain Specific Part)

Administrative Authority

City Code

Routing Domain (Building)

Area

1 2 3 4 0 0 0 5 0 0 0 0 0 1 0 4 0 0 18
Floor Machine

The sets of machines on each floor constitute somehow independent groups which consider each other neither as subordinate or as superior. Such groups we call peer groups. The address portion of a peer group is defined by the peer group mask. In example above, the peer group mask has 22 x 4 = 88 bits:

[Link].0000
Copyright C 1999 by Marko Vuskovic

2-23

NNI SIGNALING (Cont.)

Copyright C 1999 by Marko Vuskovic

2-24

NNI SIGNALING (Cont.)

Copyright C 1999 by Marko Vuskovic

2-25

NNI SIGNALING (Cont.)

Copyright C 1999 by Marko Vuskovic

2-26

NNI SIGNALING (Cont.)

Topology Aggregation To achieve scalability in a large network, the amount of topology data (nodal and link) must be reduced (summarized). For example, other peer groups should not see the internal structure of the peer group A.4. Therefore, the entire peer group can be reduced to a single node with a port for each link connected to the neighboring peer groups. Besides a reduction of complexity, the topology aggregation hides the internals of other peer groups for security reasons.

PG(A.2) PG(A.1)
A.1.1 A.4.2 A.4.3 A.2.2 A.2.1

PG(A.3)
A.4.1 A.3.4

A.4.4 A.4.5

A.4.6

PG(A.4)
Topology aggregation

Link A.4 - A.2

Port 2 Link A.4 - A.1 1 A.4 3 Link A.4 - A.3

Copyright C 1999 by Marko Vuskovic

2-27

NNI SIGNALING (Cont.)

Copyright C 1999 by Marko Vuskovic

2-28

NNI SIGNALING (Cont.)

Copyright C 1999 by Marko Vuskovic

2-29

NNI SIGNALING (Cont.)

Hiearachical View of a Node Hierarchical topologydramaticallyreduces the amount of information each node has to maintain in its data base and the amount of traffic that is flooded across the network. Suppose there are 20 PGs at the highest level, each PG at the highest level has 50 child PGs, and each child PG has 100 physical nodes. Then: the total number of physical nodes in the network is: 20 X 50 X 100 = 100,000 the number of logical nodes that a physical node can see is: 20+50 + 100 = 170

ATM Network
Hierachical view of the node A.1.1

(20) (20)

A
(50)

(20*50)

A.1
(100)

A.2

A.3

A.3

A.4

A.1.1

A.1.2

A.1.3
(20*50*100)

A.1.1. can not see nodes in the subtrees of A.2, A.3, A.4, B, C

Total number of nodes: 20*50*100 = 100,000 Maximum number of nodes which a physical node can see: 100+50+20 = 170
Copyright C 1999 by Marko Vuskovic

2-30
NNI SIGNALING (Cont.)

Copyright C 1999 by Marko Vuskovic

2-31

NNI SIGNALING (Cont.)

Hello Protocol When a switch is turned on, or when a switch is added to the network and turned on, it will send, through the Routing Control Channel (RCC), which has VPI = 0, VCI = 18, a HELLO message to all its neighbors. The HELLO message contains ATM address of the sender, its port number, and its peer group mask. Neighbors will then answer by sending similar HELLO message to the neighbor who has first sent the message. HELLO messages are restricted to the neighbors only: they contain the state information of immediate neighbors. If a HELLO message is initiated it is sent to all neighbors (at all sender's ports). If a HELLO message is an answer to the first HELLO, they are sent only to the source of the first HELLO. This process is continuous, i.e. the nodes send the first HELLO messages periodically, so the network can adapt to changing information (new nodes, existing nodes turned off, failed links, etc.). The border nodes will soon discover that they have a neighbor that doesn't belong to its peer group. This information will reach the PGL who will then establish an uplink to the higher-level peer group. This leads to the situation when each logical node eventually gets its hierarchical view of its peer group topology, including higher-level parent nodes.

1 2 RCC

Just turned on

C A
1 2 RCC

Copyright C 1999 by Marko Vuskovic

2-32

NNI SIGNALING (Cont.)

Flooding Topology State Information The purpose of the HELLO protocol is to make the neighbors aware of one's existence or nonexistence, and to collect the topology state information within a PG. The topology state information contains information exchanged through the HELLO protocol, the operational status of links, plus metrics and attributes of links and nodes: maximum CTD Cell Transfer Delay), maximum CDV (Cell Delay Variation), maximum CLR (Cell Loss Ratio), administrative weight, ACR (available cell rate), CRM (cell rate margin), etc. The gathering of this information is done through a complex process that involves HELLO messages and an elaborate hello state machine. The topology state information is organized into data base units called: PNNI Topology State Elements. A collection of PTSEs in each node forms the node topology data base. The nodes make sure that their topology data bases are consistent by checking recency of exchanged information. This process is called data base synchronization. Once a node knows that its topology data base is consistent with its neighbors, it will share the information with all members of the PG. One or several PTSEs are packed into PNNI Topology State Packets (PTSP) and sent to other nodes by flooding (see subsection Routing, Appendix C). Flooding of PTSEs is periodic, so the network nodes can always have an updated information about the network status and topology.

Call Setup and Routing As a result of the hello protocol and of the continuous flooding of PTSPs, each node can maintain the necessary routing knowledge about the entire network. This knowledge consists of the detailed topology state information about peer nodes within node's PG, and the summarized routing information about logical nodes within the ancestral PGs. Therefore each node is capable of determining the entire route, from itself to any destination node. The PNNI uses a source routing mechanism, as opposed to traditional hop-by-hop routing (as in RIP). This means that the ingress node, upon a setup request from an end station, decides about the entire route. The ingress node uses the Open Shortest Path First (OSPF) algorithm for finding the optimal route. The information about the optimal route is packed into a sequence of Designated Transit Lists (DTLs). Each DTL is a list of nodes that define the route through a single PG. With each DTL is associated a pointer which indicates the next node to be traversed. If the route crosses the PG boundary, then the additional DTLs are added to the sequence, which are related to the ancestral PGs. The sequence of DTLs is organized as a stack, with the lowest level PG at the top, and the highest level PG at the bottom. The DTL stack is passed to the next node along with the SETUP or ADD PARTY message. Each time the message moves from one node to another, the node pointer is incremented. When the message crosses into the next PG, the top DTL is removed from the stack and replaced by a new DTL which corresponds to the new PG's lowest-level node list (see example on the next page.)
Copyright C 1999 by Marko Vuskovic

2-33

NNI SIGNALING (Cont.)

Copyright C 1999 by Marko Vuskovic