Better to know some
... than all
Local Area Networks
Local Area Networks (LANs) have become an important part of most computer installations. Personal computers have been the main driving force behind the LAN proliferation. As personal computers became more widely used in office environments, so it became desirable to interconnect them to achieve two aims: to enable them to exchange information (e.g., e-mail), and to enable them to share scarce and expensive resources (e.g., printers). LANs have been so successful in realizing these aims that their cost is well justified even when there are only a handful of participating computers.
Current LANs are used for interconnecting almost any type of computing devices imaginable, including mainframes, workstations, personal computers, file servers, and numerous types of peripheral devices. Many LANs are further connected to other LANs or WANs via bridges and gateways, hence increasing the reach of their users.
A LAN consists of four general types of components:
* User station. This provides the user with access to the LAN. The most common example is a personal computer. The user station runs special network software (usually in form of a driver) for accessing the LAN.
* LAN protocol stack. This implements the LAN protocol layers. This usually takes the form of a hardware card inside the user station, containing a microprocessor and firmware which implements the non-physical protocols.
* Physical Interface Unit. This directly interfaces the user station-based LAN hardware to the LAN physical medium. The exact form of the PIU is highly dependent on the LAN physical medium. Coaxial cable connectors and cable TV taps are common examples.
* Physical Medium. This provides a physical path for signals to travel between stations. Coaxial cable, optical fiber, and infra red light are examples.
Topologies and Access Protocols
There are two general categories of LAN topologies: bus and ring.
The bus topology uses a broadcast technique, hence only one station at a time can send messages and all other station listen to the message. A listening station examines the recipient address of the message and if it matches its own address, copies the message; otherwise, it ignores the message.
The ring topology uses a closed, point-to-point-connected loop of stations. Data flows in one direction only, from one station to the next. As with the bus topology, transmission is restricted to one user at a time. When a station gains control and sends a message, the message is sent to the next station in the ring. Each receiving station in the ring examines the recipient address of the message and if it matches its own address, copies the message. The message is passed around the ring until it reaches the originator which removes the message by not sending it to the next station.
Given that access to the bus or ring is restricted to one station at a time, some form of arbitration is needed to ensure equitable access by all stations. Arbitration is imposed by access protocols. A number of such protocols have been devised:
* Carrier Sense. This protocol is applicable to a bus topology. Before a station can transmit, it listens to the channel to see if any other station is already transmitting. If the station finds the channel idle, it attempt to transmit; otherwise, it waits for the channel to become idle. Because of an unavoidable delay in a station's transmission to reach other stations, it is possible that two or more stations find the channel idle and simultaneously attempt to transmit. This is called a collision. Two schemes exist for handling collisions:
* Collision Detection. In this scheme a transmitting station is required to also listen to the channel, so that it can detect a collision by observing discrepancies in the transmission voltage levels. Upon detecting a collision, it suspends transmission and re-attempts after a random period of time. Use of a random wait period reduces the chance of the collision recurring.
* Collision Free. This scheme avoids collisions occurring in the first place. Each station has a predetermined time slot assigned to it which indicates when it can transmit without a collision occurring. The distribution of time slots between stations also makes it possible to assign priorities.
* Token Ring. This protocol is applicable to a ring topology. Channel access is regulated by a special message, called a token, which is passed around the ring from one station to the next. The state of the ring is encoded in the token (i.e., idle or busy). Each station wishing to transmit needs to get hold of the idle token first. When a station gets hold of the idle token, it marks it as busy, appends to it the message it wishes to transmit, and sends the whole thing to the next station. The message goes round the ring until it reaches the intended recipient which copies the message and passes it on. When the message returns to the originator, it detaches the message, marks the token as idle and passes it on. To ensure fair access, the token should go round the ring, unused, at least once before it can be used by the same station again.
* Token Bus. This protocol is applicable to a bus topology but makes it behave as a ring. Each station on the bus has two other stations designated as its logical predecessor and its logical successor, in a way that results in a logical ring arrangement. A special message is provided which plays the role of a token. Each station receives the token from its predecessor, readdresses it to its successor, and retransmits it on the bus. The rest of the protocol is as in a token ring.
The role of the physical layer is the same as in the OSI model. It includes the connectors used for connecting the PIU to the LAN and the signaling circuitry provided by the PIU. (The next section describes the transmission methods employed by this layer.) The OSI data link layer is broken into two sublayers.
The Media Access Control (MAC) layer is responsible for implementing a specific LAN access protocol, like the ones described earlier. This layer is therefore highly dependent on the type of the LAN. Its aim is to hide hardware and access protocol dependencies from the next layer. As we will see shortly, a number of MAC standards have been devised, one for each popular type of access protocol.
The Logical Link Control (LLC) layer provides data link services independent of the specific MAC protocol involved. LLC is a subset of HDLC and is largely compatible with the data link layer of OSI-compatible WANs. LLC is only concerned with providing Link Service Access Points (LSAPs). All other normal data link functions (i.e., link management, frame management, and error handling) are handled by the MAC layer.
LANs are not provided with a network layer (or any other higher layer) because such a layer would be largely redundant. Because the stations are directly connected, there is no need for switching or routing. In effect, the service provided by the LLC is equivalent to the OSI network layer service.
LAN transmission techniques are divided into two categories: baseband and broadband. In the baseband technique, the digital signal from a transmitting device is directly introduced into the transmission medium (possibly after some conditioning). In the broadband technique, a modem is used to transform the digital signal from a transmitting device into a high frequency analog signal. This signal is typically frequency multiplexed to provide multiple FDM channels over the same transmission medium.
Baseband is a simple and inexpensive digital technique. By comparison, broadband has additional costs: each device requires its own modem; also, because transmission is possible in one direction only, two channels typically need to be provided, one for either direction. Broadband, however, has the advantages of offering a higher channel capacity which can be used for multiplexing data from a variety of sources (e.g., video, voice, fax), not just digital data. It is also capable of covering longer distances, typically tens of kilometers compared to up to a kilometer for baseband.