1.2 Communication Devices

Communication devices have a specific place in the network and implement specific protocols at each of the layers. A host system, for example, may implement all the layers of the OSI stack to communicate with a peer host entity. A hub or repeater may implement only the physical layer to regenerate the signals. An Ethernet switch may implement only physical and data link layer functionality. A router primarily operates at the network layer, so it will typically implement the first three layers, physical, data link, and network.

This section introduces a few of the popular communications devices and software that supports them. This will set the stage for subsequent chapters in the book. During these discussions, the TCP/IP protocol suite is assumed throughout the network. xrefparanum illustrates a typical network architecture, which is used as a reference for the following discussion on hosts, switches and routers. A host on LAN1 communicates with a server on LAN 2 across a WAN (Wide Area Network) using routers and switches. This topology and the individual devices used in it will form the basis for the discussions in later chapters. Specifically, the functionality and issues related to Layer 2 switches and routers (sometimes termed Layer 3 switches in this book) will form the underlying thread for the discussions on communications systems design and implementation.

Figure 1.3: A typical network architecture.

Host Systems

A host system connected to a LAN, such as Ethernet, communicates with devices on its own LAN and can also communicate with devices on other networks, such as Web servers. A host system with a Web browser uses an application protocol like HyperText Transfer Protocol (HTTP) to access the content provided by the Web server. In this situation, HTTP runs on top of TCP at the transport layer, which, in turn, runs over IP at the network layer. In xrefparanum, the host software implements layers 1 through 5.

Layer 2 Switches

Layer 2 switches operate at the data link layer and switch MAC (Ethernet) frames between two LAN segments. They determine the destination Ethernet address from the MAC frame and forward the frame to the appropriate port. The port determination is done via a table which has entries in the form of a (Destination MAC Address, Port) pair. The Layer 2 switch constructs this table by learning the addresses of the nodes on each of its ports. It does this by monitoring traffic that the nodes send out on its LAN segment (see xrefparanum). The source address on a MAC frame indicates that the node with this address is present on the LAN segment (and port) over which the frame was received.

Note that Layer 2 switches do not examine the network layer IP address of the packets they are forwarding, as these devices switch MAC frames between two hosts on the same IP network. Layer 2 switches are an evolution of the earlier transparent bridges. Transparent implies that the hosts are not aware of the existence of the switches—i.e., they do not target frames to the switches but only to the destinations on the same network. The switch makes decisions on forwarding the frame to the appropriate LAN segment without the source node being aware of its presence. For forwarding IP packets to other networks, a router is used.

Routers

Routers operate at the network layer of the OSI model and can forward IP packets between a source and destination. In the example of xrefparanum, the host and the server with which it communicates are assumed to be on separate networks (also called subnets in the TCP/IP world). They are connected across a WAN (Wide Area Network) using routers.

Hosts send their “off-net” packets to the router, which, in turn, forwards these frames in the direction of the destination. If the destination host is directly connected, the packets are sent directly to the destination. If the destination host is not directly connected, the packets are sent to another router, from where they are directed towards the destination network, traversing, possibly, several intermediate networks. This is the scenario outlined in xrefparanum.

IP routers build tables using routing updates that are exchanged between neighboring routers. These tables are used for forwarding IP packets across a network. The format and processing of routing updates are defined in routing protocol specifications like those for the Routing Information Protocol (RIP), Open Shortest Path First (OSPF), and Intermediate System–Intermediate System (IS-IS) protocols.

Using Multiple Protocols

Protocol functions constitute a large part of the software in a communications system. Software that implements protocol functionality is often called a protocol stack. Each protocol stack component performs the protocol function while stacked on top of or below another. For example, a TCP stack sits on top of an IP stack in a host implementation. The IP stack could sit on top of a PPP (Point to Point Protocol) stack for communicating over a serial interface.

Communications devices often perform more than one function—for example, a router may also perform Layer 2 switching. Also, there may be a need for the router to communicate as an end node for management (e.g., SNMP, Telnet, HTTP) purposes. In the case of an Ethernet Layer 2 switch and router, , the device needs to implement Layers 1 and 2 (Ethernet physical layer and Ethernet MAC layer) to perform the Layer 2 switching function. To realize the routing function, the device will implement Layers 1, 2, and 3 (Ethernet physical layer, Ethernet MAC layer, and IP layer). For the end node function, there is a need for Layers 1 through 4 (Ethernet physical layer, Ethernet MAC layer, IP layer, and TCP layer), as well as the application layer (realized in protocols like HTTP).

Telecom Equipment

Routers and switches are typically used in the data communications world, where data transfer is done via packet switching. In the telecommunications (telecom) world, circuit switching is the traditional switching scheme used for voice communications. In this scenario, the end node is typically a phone which is connected to the telecom network though a wired or wireless link. The connection from the phone terminates at a local exchange or switching center.

The phone links to the network by using messages sent via a tone (analog) or messages (digital). Multiple switching centers or exchanges use these messages to establish the connection from the dialing phone to the destination phone, a process known as signaling. The exchanges communicate with each other using a separate protocol called Signaling System #7 (SS7). This protocol enables the end-to-end connection between source and destination phones.

After the connection is established, voice conversation is carried as either analog or digital information. A supervisory function, known as call control, sets up and characterizes the connection between source and destination using a set of rules. For example, the destination phone may have specified that it is not willing to accept calls from the source phone. In this case, the exchange (switch) will deny the call, since it would have been informed of this rule through SS7.

In the scenario described, the handset or telephone is a low-complexity device, while the central office switch is at the other end of the spectrum. Commercial switches like the Lucent 5ESS™ and Nortel’s DMS-100™ are extremely complex pieces of equipment which require hundreds of thousands of lines of software for call control and supervisory functions, which are the typical software components in the switch.

From an architectural perspective, circuit-switched systems can be designed such that the network contains the communications intelligence, and end systems have little or no intelligence. This is the typical view of the telecom world. The alternate is the view of the engineers in the IP/Internet world. In this view, end system devices may need to implement functions such as timeouts and retransmissions while the network performs basic forwarding functions, i.e., has little intelligence. Similar to the datagram-versus-virtual circuit argument, there is no clear winner.

Due to the increasing complexity of the software, the large circuit switching exchanges are being replaced by a new class of devices called “soft switches.” These switches separate the control processing from data or payload processing. Complex control software can reside in an off-switch workstation, rather than an embedded device like the circuit switch. This strategy increases architectural flexibility, since the workstation software can be upgraded without affecting data processing. This separation of the control and data processing functions is common among recent communications software and systems.

Phones

Traditional phones use analog signals to communicate with the central office exchange. Some cordless phones still use analog methods to communicate with the central office to record, store, and retrieve messages. Office phones usually communicate with a local exchange located within the office, which is often called a private branch exchange (PBX) and is usually a scaled-down version of the central office exchange. Often the private exchanges use a proprietary digital communications and require software functions on them built into the phones. The private exchange will, in turn, use a digital method of communication with the central office exchange using a line like a T1 or E1 line.

Cellular phones now use digital communication with the network. These phones have a set of protocols starting from the physical layer that they use to communicate with the base transmission station (BTS). The software implementing these protocol stacks includes functionality to send periodic signals to the base transmission station to indicate the location of the phone within the cellular network as well as set up connections through the cellular network to another phone. While the cellular network is quite complex, some of the complexity is on the handset itself.

Convergence

Many companies typically have two networks — a telephone network and a data network. The telephone network is a circuit-switched network both internally and externally, while email, file transfer and collaborative work is done using IP on a packet- switched network. To reduce complexity, there is a movement towards a single, integrated network, where even voice is carried over the network in IP packets. In a typical voice-over-IP implementation, a phone is connected to an Ethernet network instead of a private branch exchange. Voice is sampled and digitized into discrete packets and sent over the Ethernet network. Packets are forwarded to reach destinations, which could be other IP phones or analog phones. With analog phones, a gateway converts the packets back into analog information to communicate with the analog phone.

Simple phones of the analog world are being replaced by a complex Ethernet phone, while the relatively homogeneous circuit-switching network is being supplanted by a more complex topology, including gateways and soft switches. All of these involve a fair degree of software complexity and introduce new system requirements. On a network where voice and data traffic travel over the same links, voice traffic must be given a higher priority, since a voice packet that arrives late is meaningless to the listener. In summary, phones are increasing in complexity with a need for protocol stacks to be implemented on them.