OPTI 500 A, Fall 2011

Homework #1

Due September 21st, 2011

1. Describe the difference between time division multiplexing and wavelength division multiplexing. Use sketches, but no more than one paragraph.

2. Calculate the wavelength, in nanometers, of the center wavelength of the ITU DWDM grid.

3. List four data rates specified in the SONET protocol.

4. In your own words, describe the categories of switching used in communication networks that we have discussed in class. Use no more than two sentences to describe each category.

5. In your own words, describe the function of each of the OSI Network Model layers. Use no more than two sentences to describe each layer.

Connection-oriented NetworksSONET/SDH, ATM, MPLS and OPTICALNETWORKS

Harry G. Perros

1

Introduction

This book deals with several different circuit-switching networks and connection-orientedpacket-switching networks. These networks, although seemingly different, have all beenbuilt around the notion of a connection. That is, a connection has to first be set up betweentwo users before they can communicate. Such a connection is set up by allocating networkresources to it. The nature of these resources, as will be seen in this book, depends onthe type of the network.

The notion of a connection is also prevalent in the IP network and IP-related protocols.For instance, a TCP connection has to be set up before two TCP users can communicate.This type of connection, however, is not the same as the connection in circuit-switchingnetworks and connection-oriented packet-switching networks. For instance, let us consideran IP network that runs over Ethernet. In this case, when two peer TCP protocols set upa connection, the IP routers and the Ethernet switches are not aware of this connectionand so do not allocate any resources to it.

In this chapter, we first describe the concept of a connection as used in this book, andthen give examples of connections from the circuit-switching and connection-orientedpacket-switching networks described in this book. Subsequently, we describe the orga-nization of this book and the scope and objectives of each chapter. Finally, we presentsome of the well-known national and international standards committees involved withthe standardization process of networking architectures and protocols.

1.1 COMMUNICATION NETWORKS

Communication networks can be classified into the following two broad categories:switched communication networks and broadcast communication networks. As shown inFigure 1.1, switched communication networks are further classified into circuit-switchingnetworks and packet-switching networks. Circuit switching and packet switching are twodifferent technologies that evolved over a long time. Examples of circuit-switching net-works are the telephone network and the wavelength routing optical network. Examplesof packet-switching networks are the IP network, ATM, frame relay, and MPLS networks.Examples of broadcast communication networks are packet radio networks, satellite net-works, and multi-access local networks (such as the Ethernet).

Packet-switching networks are further classified as connection-oriented networks andconnectionless networks. Examples of connection-oriented networks are: X.25, ATM,frame relay, and MPLS. The prime example of a connectionless network is the ubiquitousIP network.

Connection-oriented Networks Harry Perros 2005 John Wiley & Sons, Ltd ISBN: 0-470-02163-2

2 INTRODUCTION

• Telephone network • Wavelength routing network

Communication networks

Circuit-switched networks Packet–switched networks

Connection-oriented networks Connectionless networks

• • •

Ethernet Packet radio network Satellite network

• X.25• ATM• Frame relay• MPLS

• IP network

Switched communication networks

Broadcast communication networks

Figure 1.1 A classification of communication networks.

In a circuit-switching network, in order for two users to communicate, a circuit ora connection has to be first established by the network. Specifically, three phases areinvolved: circuit establishment, data transfer, and circuit disconnect. These three phasestake place, for instance, when we make a phone call. Circuit establishment takes placewhen we dial a number. At that moment, the telephone network attempts to establish aconnection to the phone of the called party. This involves finding a path to the calledparty, allocating a channel on each transmission link along the path, and alerting thecalled party. The data transfer phase follows, during which we converse with the personwe called. Finally, the circuit disconnect phase takes place when we hang up. At thatmoment, the network tears down the connection and releases the allocated channel oneach link on the path. The connection is dedicated to the two users for the duration of thecall, even when no data is being sent. That is, the channel allocated on each transmissionlink along the path from our phone to the one we called is not shared with any otherphone calls. Also, in order for the call to be established, both stations must be availableat the same time.

Circuit switching is a good solution for voice, since it involves exchanging a relativelycontinuous flow of data. However, it is not a good solution for the transmission of burstydata; that is, data that continuously alternates between an active period and a silentperiod. Transmission of data only takes place when the source is in the active period.Such intermittent data transmission is typical in high-speed networks, and leads to lowutilization of the circuit-switching connection.

In packet-switching networks, information is sent in packets, which are passed throughthe network from node to node until they reach their destination. Error and flow controlprocedures can be built into the network to assure reliable service. In packet switching,two different techniques (virtual circuits and datagrams) can be used.

A virtual circuit imitates circuit switching and it involves the same three phases:call setup, transfer of packets, and call termination. In call setup, a connection path

EXAMPLES OF CONNECTIONS 3

is established between the sender and the receiver prior to the transmission of packets.This path, which all packets will follow, goes through the nodes of the packet-switchingnetwork. Each packet consists of a header and a payload. The header contains variousfields, of which one or more are used to identify the connection that the packet is asso-ciated with. This information is used to switch the packet through the network. Unlikecircuit switching, however, the channel capacity allocated on each transmission link isnot dedicated to the virtual circuit. Rather, the transmission link is shared by all of thevirtual circuits that pass through it. Error control assures that all packets are deliveredcorrectly in sequence. Packet-switching networks that employ the virtual circuit switchingtechnique are known as connection-oriented networks. Examples of such networks are:ATM, frame relay, and MPLS.

In datagrams, no connection is set up between the two users, and a user can transmitpackets whenever the user wants to. Each packet consists of a header and a payload. Theheader typically contains a number of different fields, including the source address andthe destination address. Each packet is individually routed through the network using itsdestination address. Since the packets are transferred separately, two successive packetstransmitted from the same sender to the same receiver could conceivably follow differentroutes through the network. Since each packet is routed through the network individually,a datagram service can react to congestion easier. Packet-switching networks that employdatagrams are known as connectionless networks. The IP network is a prime exampleof a connectionless packet-switching network. A packet-switching network can be eitherconnection-oriented or connectionless.

A broadcast network has a single communication channel that is shared by all of thestations. There are no switching nodes, as in circuit switching or packet switching. Datatransmitted by one station is received by many, and often all, stations. An access controltechnique is used to regulate the order in which stations transmit. Packet radio networksand satellite networks are examples of a broadcast network. The most widespread exampleof a broadcast network is the Ethernet. (Currently, the 1-gigabit and 10-gigabit Ethernet isnot used as a broadcast network. Ethernet is simply used for unidirectional point-to-pointtransmission between two users.)

1.2 EXAMPLES OF CONNECTIONS

As mentioned in the previous section, in a circuit-switching network and in a connection-oriented packet-switching network, a connection between two users has to be first set upbefore they can communicate. The connection is set up by allocating network resourcesto it. This is in contrast to the connectionless IP network, where a computer can transmitdata at any time without setting up a connection to the destination computer.

Note that connections are used in the IP network and IP-related protocols. However,these connections are logical ones between two peer protocols and do not involve alloca-tion of resources in the IP network. For instance, consider an IP network that runs overEthernet. In this case, when two peer TCP protocols set up a connection, the IP routersand the Ethernet switches are neither aware of this connection nor do they allocate anyresources to it. This of course is not the case when IP runs over a connection-orientedpacket-switching network such as ATM, as will be seen in Chapter 3. Also, in the casewhere IP is used with a diffserv, network resource allocation does take place, but just foran aggregate of connections.

4 INTRODUCTION

In this section, we give examples of connections of the packet-switching and circuit-switching networks described in this book. Specifically, we describe an ATM connection,an MPLS connection, a telephone connection, and a wavelength routing optical net-work connection.

1.2.1 An ATM Connection

Before we describe an ATM connection, we detour to examine briefly how packets areswitched in a connectionless IP network.

Let us assume that Computer A sends IP packets to Computer B, as shown in Figure 1.2.Each IP packet consists of a header and a payload, and the header contains the IP destina-tion address of Computer B. When a packet arrives at IP router 1, the header is examinedand the destination address is used in a forwarding routing table in order to find outthe next IP router to which the IP packet has to be forwarded. In our example, the nexthop router is IP router 2. IP packets arriving at IP router 2 are processed the same way.That is, the destination address is looked up in the router’s forwarding routing table inorder to identify the next hop. IP router 2 will read that the destination address of the IPpackets sent from Computer A is a local address, and it will simply send the IP packetsto Computer B. The forwarding routing table in each IP router is constructed using arouting protocol, such as the open shortest path first (OSPF).

Let us now contrast the IP procedure for routing IP packets with the scheme usedin ATM networks to switch ATM packets (commonly known as ATM cells). As will beseen in Chapter 3, an ATM cell has a fixed size of 53 bytes. Of those, 5 bytes are usedfor the header and the remaining 48 for the payload. For a user to transmit traffic to adestination user over an ATM network, user A first has to request the establishment ofa connection, as shown in the example in Figure 1.3. User A sends a SETUP messageto ATM switch 1 (to which it is directly connected). The switch calculates a path to thedestination ATM user, and then decides whether the path has enough free capacity toaccept this new connection. If it does, then the switch forwards the SETUP message tothe next switch on the path (switch 2), which in turn has to decide whether to accept theconnection, based on how much free capacity it has. If it decides that it can accept thenew connection, it forwards the SETUP message to the next switch on the path (switch3), which forwards the SETUP request to user B. The connection is established whenuser B returns a CONNECT message, which is propagated all the way back to user A.

The decision as to whether a switch can accept a new connection is crucial to theefficient operation of the network. Each ATM switch tracks all of the connections carriedthrough its switch fabric, the amount of traffic transmitted over each connection, andthe quality of service (QoS) requested by each connection. The decision to accept a newconnection comes down to whether the prospective traffic can be switched according

A B

IP Router 1 IP Router 2

IP Router 3

Figure 1.2 Routing IP packets.

EXAMPLES OF CONNECTIONS 5

ATMswitch 1

ATMswitch 2

A B

ATMswitch 3

SETUPSETUP

SETUPSETUP

CONNECTCONNECT

CONNECT

CONNECT

Figure 1.3 Successful establishment of an ATM connection.

to the requested QoS, without affecting the QoS of other existing connections. Whena connection is accepted, the switch allocates bandwidth on the outgoing link for theconnection. It stops accepting new connections when it runs out of bandwidth, or whenit reaches a certain percentage of utilization.

The user starts transmitting ATM cells once it receives the CONNECT message. TheATM cells carry two fields in the header – the virtual path identifier (VPI) and the virtualconnection identifier (VCI) – which are used to identify the connection. The ATM switchuses the combined VPI/VCI value to pass a cell through its switch fabric. Specifically,as in the case of an IP router, an ATM switch maintains a table that specifies the nexthop for each VPI/VCI value. When a cell arrives at a switch, the virtual path and virtualconnection identifiers check the table for the next ATM switch. The cell is then switchedthrough the switch fabric to the output port that connects to the next ATM switch. TheATM table is considerably smaller than an IP forwarding routing table, since it onlycontains the existing ATM connections, rather than an entire set of IP addresses.

When user A completes its transmission to B, it tears down the connection by sendinga RELEASE message to ATM switch 1. This message is propagated through the switchesalong the path, and each switch releases the bandwidth it had allocated to the connection.

As we can see, transmitting packets through the IP network is a lot simpler than trans-mitting cells through an ATM network, since it is not necessary to establish a connectionfirst. On the other hand, by establishing a connection in an ATM network, the networkcan provide QoS guarantees that are not possible in an IP network.

1.2.2 An MPLS Connection

MPLS introduces a connection-oriented structure into the otherwise connectionless IPnetwork. An MPLS-ready IP router does not forward IP packets based on the destinationaddress in the header. Rather, it forwards them based on a label that is very similar infunctionality to the VPI/VCI value carried in the header of an ATM cell.

Let us consider an MPLS-enabled IP network that runs over Ethernet. In this case, aspecial MPLS header, sandwiched between the IP header and the LLC header, is used.The MPLS header contains a label that is a short, fixed-length connection identifier. TheMPLS-ready IP router, known as a label switched router (LSR), maintains a table oflabels. When an IP packet arrives at the LSR, the label carried in the MPLS header iscross-referenced to the table of labels to find the next hop. The IP packet is then switched

6 INTRODUCTION

to the destination output port of the LSR that connects to the next hop LSR. The tablecontains labels for only the existing connections, and therefore it is not as large as theforwarding routing table in an IP router.

The procedure is similar to ATM. In order for a user to transmit over an MPLS-enabledIP network, it has to first request the establishment of a connection. This is done usinga signaling protocol, such CR-LDP or RSVP-TE. The connection is known in MPLSas a label switched path (LSP). As in the case of ATM, an LSR is aware of all of theconnections that pass through its switch fabric; therefore, it can decide whether to accepta new connection or not based on the amount of traffic that will be transmitted and therequested QoS. The LSR allocates a portion of its bandwidth to a new connection, and itstops accepting new connections when it either runs out of bandwidth or reaches a certainpercentage of utilization.

1.2.3 A Telephone Connection

The telephone network is probably the oldest connection-oriented network. A telephoneswitch, known as the central office, serves many thousands of subscribers. Each subscriberis directly connected to a central office via a dedicated twisted pair line, known as a localloop. Central offices are interconnected via time-division multiplexing (TDM) links, suchas SONET/SDH links and PDH links (i.e., T1, E1, T3, and E3).

Figure 1.4 shows two telephones interconnected via two central offices. For presen-tation purposes, let us assume that the two central offices are connected via a T1 line.Transmission on a T1 line is organized into frames, with each frame containing 24 timeslots. Each time slot is 8 bits long and carries a single voice call. The frame repeatsevery 128 µsec, meaning that a particular time slot occurs once every 128 µsec (i.e. 8000times per second). Since it carries 8 bits at a time, the total bit rate of a time slot as itcontinuously repeats frame after frame is 64 Kbps.

Transmission on a T1 line is unidirectional; that is, data is routed from central office 1to central office 2. For a bidirectional transmission between the two central offices, twoseparate T1 lines – each transmitting in the opposite direction – are needed.

In order for subscriber A to talk to subscriber B, a connection has to be first established.This connection is set up by the telephone network when A picks up the receiver and dialsthe number for the called party. A signaling protocol is used to set up a connection thatruns through the central offices that are along the path from subscriber A to subscriberB. The connection involves:

(1) a dedicated line from subscriber A to central office 1;(2) a time slot (e.g. time slot i) on the T1 line from central office 1 to central office

2; and(3) a dedicated subscriber line from central office 2 to subscriber B.

Central office 1

Central office 2

Localloop

Localloop

T1 line

Subscriber A Subscriber B

Figure 1.4 A simple telephone network.

EXAMPLES OF CONNECTIONS 7

In the opposite direction, it involves:

(1) a dedicated line from subscriber B to central office 2;(2) time slot i on the T1 line from central office 2 to central office 1; and(3) a dedicated subscriber line from central office 1 to subscriber A.

These resources are allocated to the phone call between subscriber A and subscriber Buntil one of them hangs up. A telephone connection is known as a circuit ; thus, thetelephone network is a circuit-switching network.

1.2.4 A Wavelength Routing Optical Network Connection

Optical networks are based on the wavelength division multiplexing (WDM) technology,which combines multiple wavelengths onto the same optical fiber. A wavelength is afrequency on which a data stream can be modulated. Each wavelength, therefore, is aseparate transmission channel. Transmission over a WDM fiber requires W-independenttransmitters. Each transmitter is a light source (e.g. a laser), and is independently modu-lated with a data stream. The output of each transmitter is an optical signal on a uniquewavelength: λi, i = 1, 2, . . . , W . The optical signals from the W transmitters are com-bined into a single optical signal at a wavelength multiplexer and transmitted out onto asingle optical fiber. At the receiving end, the combined optical signal is demultiplexedinto the W individual signals, and each one is then directed to the appropriate receiver,where it is terminated and converted to the electric domain.

A wavelength routing optical network consists of optical cross-connects (OXCs) inter-connected with WDM fibers. An OXC is an N × N optical switch, with N input fibersand N output fibers. Each fiber carries W wavelengths. The OXC can switch optically ;that is, all of the incoming wavelengths of its input fibers are switched to the outgoingwavelengths of its output fibers without having to convert the optical signal to an electri-cal signal. For instance, the OXC can switch the optical signal on incoming wavelengthλi of input fiber k to the outgoing wavelength λi of output fiber m.

A wavelength routing network is a circuit-switching network. That is, in order for a userto transmit data to a destination user, a connection has to be first set up. This connection isa circuit-switching connection, established by using a wavelength on each hop along theconnection’s path. For example, let us consider that two IP routers (router A and routerB) are connected via a three-node wavelength routing network (see Figure 1.5(a)). The

(a) A three-node wavelength routing network

(b) A lightpath between Routers A and B

OXC 1 OXC 2 OXC 3Router A Router B

OXC 1 OXC 2 OXC 3Router A Router B

l1,..,lW l1,..,lW l1,..,lW l1,..,lW

l1 l1 l1 l1

Figure 1.5 A lightpath.

8 INTRODUCTION

link from router A to OXC 1, OXC 1 to OXC 2, OXC 2 to OXC 3, and OXC 3 to routerB is assumed to be a single fiber with W wavelengths, referred to as λ1, λ2, . . . , λW . Datais transmitted only in one direction: from router A to router B. Another set of fibers (notshown in Figure 1.5(a)) has to be used in order to transmit data in the opposite direction(i.e. from router B to router A).

Assume that IP router A wants to transmit data to IP router B. Using a signalingprotocol, A requests the establishment of a connection to B. The connection betweenrouters A and B is established by allocating the same wavelength, say wavelength λ1, onall of the links along the path from A to B (i.e., links A to OXC 1, OXC 1 to OXC 2,OXC 2 to OXC 3, and OXC 3 to B). Also, each OXC is instructed to switch λ1 through itsswitch fabric transparently. As a result, an optical path is formed from router A to B, overwhich data is transmitted optically. This optical path is called a lightpath, and it connectsrouters A and B in a unidirectional way from A to B. In order for B to communicate withA, a separate lightpath has to be established in the opposite way over a different set offibers which are set up to transmit in the opposite direction.

1.3 ORGANIZATION OF THE BOOK

In this book, we explore two connection-oriented packet-switching networks: ATM net-works and MPLS-enabled networks. ATM is a legacy network that was developed in thelate 1980s and early 1990s. It is used in the backbone to transport IP traffic, in accessnetworks such as ADSL-based networks and ATM passive optical networks (APON), andin cellular telephony. The MPLS architecture can be seen as an extension of ATM, andit can be used to introduce QoS in IP networks.

Two circuit-switching networks – SONET/SDH and optical wavelength routing net-works – are also presented in this book. SONET/SDH has been around for along time,whereas optical wavelength routing networks are relatively new. SONET/SDH is theunderlying transport network of the telephone system. It is also used in all modernpacket-switching networks, such as IP and ATM. Wavelength routing networks are alsocircuit-switching networks since the transmission of data is done using optical circuit-switching connections, known as lightpaths. We also present a new optical networkingscheme, which has not yet been standardized, known as optical burst switching (OBS).This type of optical network can be seen as lying between packet switching and cir-cuit switching.

Finally, the book contains a chapter on access networks, such as ADSL-based networks,cable modems, and passive optical networks, and a chapter on voice over ATM and voiceover MPLS.

The book consists of twelve chapters, which cover the following topics:

• Chapter 1: Introduction• Chapter 2: SONET/SDH• Chapters 3, 4, and 5: ATM networks• Chapters 6 and 7: MPLS• Chapters 8, 9, and 10: Optical networks• Chapter 11: Access networks• Chapter 12: Voice over ATM and MPLS