Typically bridges, routers, switches, and gateways are used for connecting LANs together via higher speed circuits, more commonly fi ber optical cables. A bridge is a data link layer device connecting similar types of LANs (i.e., the same data link layer protocol). Bridges have lost their place to Layer 2 (data link layer) switches as the latter became cheaper and more powerful. A router operates at the network layer and connects LANs with different data link layer protocols, but with the same net- work layer protocol. A router processes network layer messages and prepares new data link layer messages for outgoing packets. When a router receives a packet, it reads the destination address of the network layer, chooses the “best” route for the packet (via routing tables), and then sends it out to the next node on the selected route. Gateways also operate at the network layer and connect LANs with different data link layer and different network layer protocols. Figure 3.4 shows a campus network where LANs in different buildings are connected to each other via routers. This way each LAN has its own subnet designation.
The architecture of Figure 3.4 is easier to manage but it tends to impose more delays than bridging due to the Layer 3 processing. An improvement could be to
Hub Hub Hub HUB Hub Hub HUB LAN (PCs) Router Router Router ROUTER Router Router Router ROUTER Hub LAN (PCs)
connect each hub of the LANs to a switch via a separate circuit (and get rid of all the routers). In this improved confi guration, there will be more cables, but less devices to worry about and due to the switched operation, the performance will improve signifi cantly. The down side is that, the use of a single central switch introduces a reliability problem: when the switch fails, the whole network goes down. We could improve the confi guration even further by using a higher performance and a more intelligent switch. In this confi guration, we can connect all the computers directly to the switch (even eliminating the hubs). This new confi guration allows us to create virtual LANs by creating virtual LAN segments via software and assigning comput- ers to them. This way, we can assign any computer to any segment regardless of the location of the computer. This confi guration provides a more fl exible network man- agement in terms of creating project groups and assigning resources to them. Multi- switch virtual LAN confi gurations are also possible by using several switches. We should note here that this is a more complex and costly network confi guration and it is typically used for larger campus networks.
The LANs and campus networks discussed above are mostly data centric. With additional software and/or hardware changes in the end-user equipment and in the network elements, these networks can transmit voice and video as well. Typically, Voice over IP (VoIP) and multimedia-specifi c protocols and features that rely on a common network layer protocol, IP, are used to accomplish this. Separately, a PBX (a.k.a., private automatic branch exchange, PABX) can be used to connect all tele- phones in an enterprise in a campus. With one or more PBXs, a voice only campus network can be deployed. Of course, the PBXs need to be connected via trunk lines to a carrier network in turn to receive and make out-of-campus calls.
With a traditional PBX here, an enterprise need to maintain separate voice and data networks resulting in higher operations and capital expenses for the enterprise. A latest trend in the PBX development is the IP PBXs, which can switch calls between VoIP on local lines. This makes it possible to have a single line to each user for data access, as well as VoIP communications and traditional telephone communications.
MANSAND WANS
If the computers and other network elements are dispersed in a metropolitan area spanning from 3 to 30 miles, the network is typically called a MAN. These types of networks typically connect campus networks and LANs together. On the other hand, if the area spans beyond the metropolitan neighborhood to several hundred miles covering a province, a country, or even across countries, the network is called a WAN. Other than the distance differences, MANs and WANs have generally similar characteristics, so in this section, we will not distinguish between the two. Typically, these networks are built by using dedicated circuits leased from common carriers. A WAN containing the dedicated circuits is called a dedicated circuit WAN. How- ever, more typically, the computers in such a network are connected by using the communications services provided by common carriers. The services provided by common carriers can be classifi ed as circuit-switched services and packet-switched
In a dedicated circuit WAN, the circuits connecting the locations of an organiza- tion are leased from common carriers, which charge a monthly fl at fee that depends on the capacity and length of the circuit. The line is dedicated to the customer with the rights of unlimited use of the circuit. The T-Carrier services are the most com- monly used dedicated digital circuits in North America. (In Europe and elsewhere, a similar system called E-Carrier service are used.) Commonly used T1 circuits provide 1.544 Mbps data rate (equivalent of 24 voice channels, 64 Kbps data rate). T3 circuits offer 44.376 Mbps data rate (28 T1 lines). For higher data rates, common carriers offer dedicated circuits based on the Synchronous Optical Network (SONET) technology, which is an ANSI standard in the United Sates for optical fi ber transmis- sion in Gbps range [similar to ITU-T-based, Synchronous Digital Hierarchy (SDH)]. Hierarchy of data rates in SONET starts with OC-1 (optical carrier level 1) at 51.84 Mbps. Each succeeding SONET hierarchy rate is defi ned as a multiple of OC-1. For example, OC-3’s data rate is 155.52 Mbps, and OC-12’s data rate is 622.08 Mbps. Larger networks handling heavy traffi c may prefer to use OC-192 pro- viding almost 10 Gbps data rate. There are special equipments, such as Channel Service Unit and Data Service Unit, which need to be installed at the end of each dedicated circuit. Then the customer uses routers and switches to connect its loca- tions together to form a network owned and maintained by the organization itself.
Network designers must determine the best architecture that fi ts the application at hand, juggling among various factors such as delay, throughput, reliability, and the cost. There are a number of different ways of connecting the locations via dedicated lines. The ring, star, and mesh topologies are the basic dedicated circuit architectures that are used more commonly. Figure 3.5 shows these there basic topologies.
The ring and star architectures are most cost-effective since they result in less dedicated circuits to lease, whereas the mesh architecture is more costly since it requires many more circuits. As for the performance (throughput, delay), the mesh is the best, then the star, and the ring is the worst. From the reliability point of view, again the mesh architecture is the best, and the ring architecture is the worst since the network relies on a central location. To bring the down the cost of a mesh
Ring Star Mesh
architecture, a partial mesh architecture where only certain pair of nodes are con- nected directly can be used. In a partial mesh architecture, any node is reachable from any other node (i.e., the nodes that are not directly connected communicate with each other through other nodes).
Enterprises that cannot afford to put together their own dedicated circuit-based network rely on switched services provided by the common carriers. In a switched WAN, end-user equipment are connected via temporary, not dedicated, connections for the duration of the call, or session. Once the session is fi nished, the connection is no longer available for this end-user and the same connection may be offered to another user. The connection services are offered by common carriers. As discussed earlier, two types of switched services are common: circuit-switched services and packet switched services.
The circuit-switched services approach is an old way and perhaps the simplest way to have a WAN. Common carriers offer two types of circuit-switched services: Plain Old Telephone Service (POTS) and Integrated Services Digital Network (ISDN). In POTS, end-user equipments are connected to each other via telephone lines. A computer by using a modem in a location dials the number assigned to a remote computer and the telephone network provides a temporary circuit for this communication. When the session is completed, the circuit is disconnected, and may be allocated to another session. (The circuit is said to be switched to another conver- sation, or to another session.) The ISDN-based circuit-switched services gives the capability of combined transmission of voice, video, and data over the same digital circuit. ISDN services include the Basic Rate Interface (BRI), which provides 144 Kbps data rate and the Primary Rate Interface (PRI) with 1.5 Mbps data rate. ISDN services require a special modem connected to the end-user equipment. With a wider availability of packet-switched services such as the Internet, circuit-switched services are no longer attractive in WAN applications.
An enterprise wishing to use a packet-switched service fi rst leases a short connec- tion from each of its locations to the nearest Point of Presence (POP) of the service provider. An end-user equipment in the enterprise is required to break its message to be transmitted into smaller segments, called packets and attach the address of the destination equipment. Unlike the circuit switching and/or private lines, no circuit is dedicated to the two communicating parties during the communication. The service provider network simply fi nds the best route by employing the network layer routing protocols to deliver individual packets to the destination effi ciently and reliably. There are a number of packet-switching technologies and associated standards used in deploying this kind of network. X.25 is the oldest standardized packet-switched protocol; it was standardized by ITU-T (CCITT at the time). X.25 specifi es a three- layer protocol suite for the network, data link, and physical layers. Frame relay is another technology that operates at rates higher than those of X.25 up to 45 Mbps, by taking advantage of the improvements in the reliability of transmission over fi ber optic lines so that some Layer 2 in error controls could be done at the endpoints instead of the intermediate links. Another technology that the industry spent consid- erably effort to standardize and implement is called the Asynchronous Transfer Mode (ATM). A major distinguishing characteristics of the ATM technology is the use of fi xed-length packets, 53 byte “cells” instead of variable packet length approach
of all other protocols, including the Internet. Small fi xed-length cells were believed to facilitate the adoption of ATM for realtime voice transmissions, by avoiding the need for echo cancellers at the access circuits, thereby reducing the cost. The ATM also provides capabilities to enable setting of precise priorities among different types of transmissions (i.e., voice, video, and e-mail). ATM services can be provided at the same rates as SONET: 51.8, 466.5, 622.08 Mbps. With the popularity of the Internet, the ATM-based packet-switched services never became popular because of the large overhead in transmission. The technology is now mainly used in the core infrastruc- ture of major network providers. As alluded to in the above statement, the Internet is the most commonly used packet technology today. Later in this section, the Internet will be discussed in detail.
There are also some packet-switched service offerings based on the Ethernet/IP
Packet Network technologies, which are based on the Gigabit Ethernet fi ber optic
networks (bypassing common carrier network). With these services, there is no need to translate LAN protocol (Ethernet/IP) to the protocol used in MAN/WAN services. As the number of services requiring packetized 10-Gbps pipes increases (e.g., video distribution), it is anticipated that within the coming few years there will be a need for a comprehensive network solution to aggregate traffi c at rates of N × 10 Gbps. The current consensus is to standardize two such rates: N = 4 and N = 10. As discussed before, the standardization activities for carrier class Ethernet interfaces are taking place mostly in the ITU-T study group 15. [Remember that the IEEE 802.3 Higher Speed Study Group (HSSG) is focusing on campus and local area spe- cifi c concerns.] The scope of activities in the ITU-T corresponds unequivocally to a platform innovation: defi nition of a new transport container for optical transport units (OTU)/optical data units (ODU)—denoted as OTU4/ODU4—for 100-Gbps Ethernet, defi nition of the characteristics of the interface (e.g., the forward error con- trol scheme), and so on. The two alternative trajectories for the next generation Ethernet networks are summarized in Table 3.3.
In the following subsections, we discuss in more detail two popular examples of a WAN: the Internet and wireless mobile networks.
TABLE 3.3
Alternative Standardization Trajectories for the New Generation of Ethernet Networks
ITU-T SG 15 IEEE 802.3
Intended rate (Gbps) 100 40/100
Value chain Carrier networks (e.g., long distances; high reliability) Server interconnects (e.g., short distances) Technologies to
be standardized
New container for OTU/ODU, new modulation schemes, optical interfaces, network interface controllers, etc.
Extending existing technologies Characteristics of
innovation
The Internet
The Internet is a network of networks; an interconnection of thousands of LANs, campus networks, MANs, and WANs together to form a worldwide area network. Later in the protocol sections of the chapter, we will discuss a number of basic con- cepts (packet switching, IP/TCP protocols, routing, and addressing, etc.) that are used in the Internet. Here, we discuss its unique architecture, its access technologies, and future of the Internet.
To constitute the Internet, thousands of networks are connected together based on a hierarchical structure. Individual private networks and computers belonging to individual organizations and people are connected to an Internet Service Provider (ISP), and many ISPs are connected to each other via bilateral agreements and con- nections. Some ISPs are small and local or regional and some are larger and nation- wide. National ISPs provide services to their individual customers and sell access to regional ISPs and local ISPs. Regional ISPs, connected with National ISPs, provide services to their customers and sell access to local ISPs. Finally, local ISPs sell access to individual organizations or residential customers. More formally, ISPs are classifi ed into three tiers: Tier 1 ISPs, the largest ones; Tier 2 ISPs, the ones that buy connectivity from Tier 1 ISPs; and Tier 3 ISPs, which buy connectivity from Tier-2 ISPs. Figure 3.6 shows the basic hierarchical architecture of the Internet today. There is a payment-compensation scheme established by the ISPs. ISPs at the same level usually do not charge each other for exchanging messages. This is called peering. Higher-level ISPs charge lower-level ones: Tier 1 ISPs charge regional Tier 2 ISPs, which in turn charge Tier 2 ISPs. Of course, Local ISPs charge indivi- dual residential and corporate customers for access. Based on this classifi cation, a Tier 1 ISP can reach every other network on the Internet without purchasing connection from any other network. However, a Tier 2 network may do peering with some other networks, but still must purchase connection from a Tier 1 ISP to reach
Tier 2 ISP Tier 2 ISP
Tier 2 ISP Tier 3 ISP
Tier 3 ISP
Tier 3 ISP Tier 3 ISP Tier 3 ISP
Tier 3 ISP
Tier 2 ISP
Tier 2 ISP Tier 2 ISP
Tier 3 ISP Tier 3 ISP
Tier 1 ISP
Tier 1 ISP Tier 1 ISP
Tier 3 ISP
some points of the Internet. A Tier 3 ISP must purchase connections from other ISPs to reach the Internet.
As one can observe, there is no one single company in charge of the Internet. However, the Internet Society (ISOC), an open membership professional society with over 175 organizational and 8000 individual members in over 100 countries, provides a platform to develop various industry standards and agreements that are used in the Internet. Among other things such as public policy and education, ISOC is also involved in standards development through its IETF to develop Request For Comments (Internet standards); Internet Engineering Steering Group to manage the standard process; Internet Architecture Board to provide strategic architectural over- sight; and Internet Research Task Force to focus on long-term research issues involv- ing the future of the Internet.
To access the Internet, telephone lines via dial-up modems and T1/T3 lines were commonly used until recently. However, today, the so-called broadband access tech- nologies, which include cable modems and Digital Subscriber Line (DSL) modems that provide higher data rates in the Mbps range, are more common. DSL is a tech- nology designed to provide high-speed data transmission over traditional telephone lines. A DSL modem needs to be connected at the customer site to the customer’s computer and a corresponding circuit must be part of the central offi ce switch at the phone company’s offi ce.
There are several types of DSL technologies: Asymmetric DSL (ADSL) uses a simplex data channel for downstream traffi c and a slower full-duplex data channel for upstream traffi c (data rates for downstream range between 1.5 and 9 Mbps and for upstream the range is between 16 and 640 Kbps); Very High Speed Digital Subscriber Line (VDSL) is designed for local loops of 1000 ft, ideal for video trans- mission (data rates over 100 Mbps are offered by the latest VDSL products). As mentioned above, another technology for broadband access to the Internet is based on the coaxial cable that cable TV companies use to bring the TV channels to the residential buildings. A modem, called the cable modem, is used to modulate signals coming from the PC so that the transmission can take place on the cable TV wires. In this case, cable TV companies become ISPs.
The most common protocol used by the cable modems is called the Data Over Cable Service Interface Specifi cations (DOCSIS) produced by CableLabs, an indus- try forum founded by the cable TV companies. The data rates for the cable modem technology are much higher than those of DSL technology: downstream: 27–55 Mbps; upstream: 2–10 Mbps. The infrastructure is such that the users in a neighborhood share a multipoint circuit (300–1000 customers per cable segment). This means that all messages on a circuit are heard by all computers on that circuit. This also means that the bandwidth is shared among the users in a segment (if a large number of people in the neighborhood use the Internet at the same time, the degradation of the service will be noticeable). To handle the termination of the circuits and provide demodula- tion, the Cable Modem Termination System (CMTS) is used for the upstream traffi c. The CMTS converts data from DOCSIS to the IPs. For the downstream traffi c only, a Combiner is used to combine the Internet traffi c with the TV video traffi c.
In addition to the DSL and cable modems, there are a number of other techno- logies, some old and some new. We mention, among others, the satellite, wireless
fi xed access, and wireless mobile access, as some that have been in use. As for the new and upcoming technologies, we should mention the Passive Optical Networking (PON), also called Fiber to the Home (FTTH), already being deployed aggressively by the telephone companies. The PON, using Dense Wavelength Division Multiplexing