Local Area Networks

A Local Area Network cannot be defined without first defining networks. Therefore, a network is defined as 'an interconnected collection of autonomous computers'. This means two things, first that no computer can control another computer directly and no computer relies on another for its own operation. Second and most important, that they can exchange information. Networks are classified according to their size, often a network will start small and grow with the organization using it and its needs.

A LAN is the smallest, usually consisting of 3 to 50 nodes located within the same building or organization. It operates at speeds between 1 and 100 mega bits per second (mbps), this is mainly due to the relatively short distance between nodes (1m to 10 km). Also because of the short distances, error rates are very low.

Two or more LAN segments can be connected together using routers to create an internetwork. This happens when several LANs exist and we wish to connect them. A Metropolitan Area Network (MAN) or Wide Area Network (WAN) provides access to more distant resources than can be accessed using a LAN. A MAN is smaller than a WAN, usually covering a city or several departments scattered on a large area. It usually consists of high speed special "backbone" cable (like optical fibers) or wireless communications (like microwave dishes pointed at each other). This cabling and/or equipment is usually provided by the organization owning the network. A WAN on the other hand, provides connections countrywide or even worldwide using telephone lines or satellite connections. These are usually leased from governments. Because of the distances and carriers involved, error rates are high and transmission speeds are low (1 mbps) in comparison to LANs. Fortunately, typical traffic on a WAN is usually much lower than on a LAN.

It must be emphasized that the previous definition of LANs, MANs and WANs in the light of recent technological advances may be misleading. A clear distinction cannot be made, as the technology advances, limitations on each kind become non-existent. What we end up with usually a mixture of the three types becoming the operating network.

This may seem like a useless question, but it is not. A network can offer many advantages. The following highlights these advantages.

Program and File Sharing : Using a network version of a specific program, several users can use the same program without having to buy and install individual versions for each one. Also all the data files are stored in one location (file server) available through the LAN. This eliminates the need for diskettes and the time to take it from one computer to the other.

Network Resource Sharing and Economical Expansion : Devices like printers, plotters, modems, faxes and storage space that a single user may not need all the time can be shared and used by all the users. This eliminates the need for individual devices and thus drives costs down.

Database Sharing : Databases are ideal for networks. A user always accesses up-to-date information. Using record locking, multiple users can access the same record while a single user (with editing access or rights) may edit it. This ensures that data is not corrupted.

Electronic Mail : This function lets users communicate easily. Messages are dropped in electronic "mailboxes", to be retrieved by the recipient whenever he wishes to.

GroupWare : This is a new class of software enabling users to interact and coordinate their activities without leaving their desks. A project group may be created and given access to special resources. This allows people from different locations (possibly separated by thousands of kilometers) to work on the same project.

Centralized Management : Servers and shared resources may be all gathered in one area making management tasks easier (hardware and software upgrading, regular backups, maintenance...).

Security : Each user must first login to his account using a password. Each account can be given access to authorized areas in the network depending on the user.

Multiple Operating System Access : Users with different Operating Systems can all access the network and exchange information as long as the Network Operating System supports it.

Enhancement Of The Corporate Structure : Users in a specific department need not be in the same physical location. Connection to their department and their manager is done through the network. This is particularly useful when people from different areas have to work together for a specific project. Instead of relocating their offices, they would just communicate and work together through the network.

LAN Applications

The automated office local network can provide several services to improve productivity. The main feature of the office's LAN is resource sharing. The network provides services for office automation functions such as: document distribution, electronic mail, electronic filing (DBMs storage), word and text processing, remote printing, digital facsimile and voice storage.

Several Japanese companies, the American National Standards Institute (ANSI) and the LAN standardization project team in U.K. had presented papers on the Integrated Services Local Network (ISLN). The possible integration of speech, data and video so that they are carried by one transmission medium and switched in one switch could lead to exciting improvements in the field of office automation.

LANs can aid factory automation, including automated manufacturing techniques such as CAD (Computer Aided Design), CAM (Computer Aided Manufacturing), robotics and numerically controlled manufacturing processes.

LAN can also provide a communication facility for exchanging data between the manufacturing sites and the rest of the corporation's departments. Also, LAN can support distributed process control on a real time basis, thus large tasks can be distributed among stations for better performance and reliability.

A LAN in a bank can provide faster and better services for customers. A customer's transactions can be handled more efficiently and faster. Branch managers can access the details of accounts' information via their administrative terminals, when trying to answer a customer's questions and resolve any problems. Automatic Teller Machines (ATM) can be easily connected to the LAN to allow customers to perform simple transactions after normal business hours and on weekends.

The Topology of a network determines how nodes are connected together. It also influences the choice of media, hardware and protocols used. Since it determines how the wiring of the LAN is to be done, it may often be the hardest obstacle to overcome. Several factors influence the choice of topology to be used. The most important of which are Cost, Flexibility and Reliability. The transmission medium has to be physically installed in the building. In an ideal case, it would be carried out before the building is occupied and should be able to cope with foreseen grof the network. In a more practical situation, ducts, raceways and wiring closets would have to be installed (which means more cost). Some nodes may need to be moved or added at specific locations due to the change of the office layout, topologies with high flexibility allow this task to be carried out easily. Last of all, Reliability, a failure can always occur, the topology should allow the location of the fault to be easily detected and isolated quickly with minimal side effects to the remainder of the network.

Many topologies are possible, but only three have had a major role in LANs, as they fit the above criteria to a certain extent. These are the Star, Bus and Ring Topologies.

In this topology all the nodes in the network are connected to a single node called the central node. This is the topology most widely used for networks dealing with data processing or voice communication. Usually wiring closets are used to group connections to a single area. This provides flexibility to this topology as to allocating nodes within that area, without having to go back to the central node each time. The Star is not normally used in a LAN, but rather in hybrid topologies. Recent computer systems tend to rely less (or not at all) on the host computing power. This has led to a falloff in the use of this topology. However, it is the dominant topology in traditional communications, so it will probably stay in use for some time.

The advantages of this topology are :

• Ease of Service : All the connections meet at concentration points, either at the central node or at intermediate wiring closets. This provides easy access for administrative tasks.
• One Device Per Connection : If a failure occurs, only one connection gets disconnected and the rest of the network is not affected. It is very simple to disconnect the failing node from the network.
• Centralized Control/Problem Diagnosis : Since every node on the network is directly connected to the central node by a single cable, faults are easily detected and isolated until they are fixed.
• Simple Access Protocols : Any connection involves only the central node and a node. In this case, sharing the medium is not a problem, because each node has its own connection. Access Protocols are discussed later under Access Control.

• Long Cable Length : A cable must run from the central node to each node on the network. While the cost for the cable is small, the problem typically occurs with too many cables inside one cable duct making installation and maintenance difficult and costly.
• Difficult to Expand : To add a new node, a connection has to be made all the way to the central node. Usually, some redundant connections are added at the initial wiring to allow for expansion, but problems can always occur due to unforeseen expansion.
• Central Node Dependency : While failure in a single node does not affect the network, failure in the central node shuts down the entire network. This requires the central node to be highly reliable and some redundancy must exist in this node.

A popular topology for data networks is the Bus. It consists of a single length of cable. All the nodes get attached to this cable. This configuration is also known as multidrop line. It is widely used in data communications where the host is at one end and the terminals are connected at the other end.

The Advantages of the Bus are :

• Short Cable Length and Simple Wiring Layout : This topology offers the shortest cable length as a single common path exists between all nodes in the network. Installation costs and maintenance are decreased as a result.
• Reliable Architecture : The Bus is inherently simple which makes it very reliable from a hardware point of view. The data propagates along one single cable to which all the nodes are connected.
• Easy to Extend : New nodes can be added anywhere along the cable. If the cable is not near enough to a new node, an additional segment can be added using a repeater as shown in figure.

• Fault Diagnosis is Difficult : Even though the Bus is reliable, faults can occur. Fault isolation tends to be difficult, because control of the network is not centralized. Fault diagnosis may have to be performed from several nodes.
• Fault Isolation is Difficult : If a node is faulty, it must be disconnected at point where it connects to the network, this is a relatively simple process. If the medium is at fault, however, an entire segment of the network has to be disconnected.
• Repeater Configuration : When a Bus-type network gets extended using repeaters, reconfiguration may be necessary. This may involve changing cable lengths, adjusting terminators, etc...
• Nodes Must Be Intelligent : Due to the lack of a central node, the individual nodes must manage access control. This approach tends to increase costs, whether it is solved by hardware or software.

The third and final "pure" topology used is the ring or loop. In this topology, each node is directly connected to two neighboring nodes and so on until the ring is closed. Data is received from one node and transmitted to the next one, this means that data flows in only one direction. After the transmitted data passes through each node in the network, it returns to the transmitting node, which removes it. It should be noted that data passes "through" and not passes "past" each node. Each node receives the signal and retransmits it, thus amplifying it in the process if it had somewhat deteriorated. The receiving node can mark the data as read before retransmitting it, providing a way to inform the sender that the message was received.

• Short Cable Length : The length of cable used is slightly longer than that of a Bus and small relative to that of a Star. As for the Bus, this increases network reliability.
• No Wiring Closets : They are not needed at all, because there is only one cable going from one location to another. This in turn reduces cost and makes more space available for other uses.
• Suitable for Optical Fibers : As will be discussed later, optical fibers are essentially uni-directional, which makes them very suitable for the ring. Also, different media types can be used for different parts of the network, because each node receives the signal, then retransmits it, as was highlighted before. A segment of optical fibers can be used where noise is a problem for example while using copper media through the rest of the network.

• Node failure causes network failure : Since the data transmitted must go through every connected node, if a failure occurs in one node, the entire network fails. No transmission can occur until the failed node has been removed from the network.
• Difficult to diagnose faults : When one node fails, the entire network is affected, making the fault difficult to find. Several segments have to be examined before finding the fault.
• Network Reconfiguration is Difficult : It is difficult to extend the network, or modify its layout. For instance, it is not possible to shut down a segment while maintaining the rest of the network.
• Topology affects the access protocol : Each node receives all the data going through the network. There is a "responsibility" on each node to pass on the data it received. The access protocol must take this into consideration.

As we have seen, each "pure" topology has its own advantages and disadvantages. By combining some of them together, a more useful network can be obtained. These are called Hybrid topologies. In addition to these hybrid topologies, one can always create several LANs of different topologies and connect them together using bridges and routers.

Only two hybrid topologies will be discussed as examples of what can be achieved using hybrid topologies. The first is called the tree. It is mainly a bus shaped like a tree with a central root branching and sub-branching. The difference between the tree and a Bus with repeaters, is that the "root" receives the transmitted information and retransmits it through the whole network. This eliminates the need for repeaters. The advantages and disadvantages are similar to the Bus, with some extras.

Two main advantages over the Bus are that it is easier to extend by adding new branches. Also Fault isolation is easier, since it is possible to disconnect whole branches from the network without affecting the entire network.

The main disadvantage is a dependency on the root. If it fails, the entire network fails, this makes the tree similar to the star in this respect.

The other hybrid topology being discussed is the star-ring. This topology tries to achieve the best of both. The layout looks like a number of concentration points connected in a ring. From each concentration point, several nodes can be connected. Electrically, the star-ring operates like a pure ring, because each node is connected to two other nodes. The difference is in the wiring which is arranged like a star. The star-ring is sometimes called the star-shaped ring because of this.

There are several advantages over those of the pure ring. First, fault diagnosis and isolation is easier because of the presence of concentration points. The ring connecting the concentration points is smaller than the total ring, making diagnosis and isolation manageable. Second, expansion is easy, each concentration point can have extra wiring for additional nodes. The following step being adding a new concentration point for handling more nodes. Third, the cabling is easier due to concentration points. The ring cuts down the number of wires in a cable duct, while the star simplifies the installation process.

The disadvantages are few and easier to overcome. First, depending on the implementation, the concentration points may have to be intelligent to assist in networking tasks, fault detection and adding new nodes. Second, the cabling although an advantage is also a disadvantage. The ring between concentration points is critical, which may require redundant cables as back-up. Also as the larger section of the ring is laid out in a star shape, a considerable amount of wiring is needed.

Choosing the correct media for the network can be a tough choice. Cable cost is also an important factor. Installation problems can further complicate the choice. Wireless LANs are always an attractive option.

There are several types of cable each with its own advantages and disadvantages. A comparison table is included for convenience.

Consists of two identical insulated wires of copper braided together. Twisted Pair cable is the cable used for public telephone applications. For data communications, Data Grade Medium (DGM) must be used, these are higher quality cables than the normal cable. Its main advantage is that is cheap, easy to install (sometimes already installed) and simple. However, it is susceptible to outside noise and has length limitations because of high attenuation.

Consists of a solid wire core surrounded by one or more foil or braided wire shields, each separated from the other by an insulator. The signal is carried on the inner core while the shield provides the ground. It is less prone to noise and interference than twisted pair cables and also has better bandwidth, higher transfer rates and lower error rates. An additional shield can be added to reduce interference. Unfortunately, it often acts like an antenna, enabling intruders to monitor transmissions and the wire itself may pick up unwanted interference.

An optical fibers is a thin strand of glass or glass-like material (core) surrounded by a concentric layer of glass (cladding). Data signals are transmitted in the form of modulated light beams. It is not prone to noise or interference, nor does it emit any. Very high bandwidth and transmission rates are allowed on optical fiber while suffering from very low (or no) error rates. It is however the most expensive type.

Copper has been used for many years, it is a well understood media, both in manufacturing and maintenance. A lot of supporting technology has been developed to assist it. It is also very cheap because of its widespread. Optical media on the other hand is still relatively new and subject to research and development. A major problem with optical media occurs when trying to connect two sections together. While this is straightforward with copper, it is not as easy with optical fibers. Optical fibers are maturing fast however and it should be expected that it will play a major role in years to come.

An important advantage of optical fibers over copper is that attenuation tends to be uniform over a larger range of frequencies, allowing for more bandwidth to be used. It also does not suffer from noise, making it a very reliable media. As for security, it is very difficult to tap into an optical fiber without corrupting the medium and as no signal is being radiated, the signal cannot be intercepted as can be done with copper media.

Wireless media is very attractive because there is no cable cost or maintenance. There are several ways to implement this method. For instance, infrared light may be used, it offers a wide bandwidth, high transmission rates. The problem with infrared light is that it operates on line of sight. If the source and destination do not "see" each other, they cannot transmit. Because infrared light can also come from other sources, interference or distortion may occur. Another way is to use narrow-band also called radio link, each node tunes in to a certain frequency and sends and receives on it. The signal can penetrate physical obstacles such as walls. Problems occurring are the same for radio transmissions, like interference from a nearby channel, a station not precisely tuned in or even a stronger station operating at the same frequency. One more choice is present in spread spectrum radio, where the signal is spread over a range of frequencies. A code is used to spread the signal and the receiving station uses the same code to retrieve it. Interference is less of a problem and several signals can overlap without problem.

In baseband transmission, the signal to be sent is directly applied to the physical medium. As the signal propagates it is attenuated causing the quality of the signal to decrease with the distance traveled. This usually limits the transmission speed and maximum distances.

A signal can be easily modulated into a carrier using amplitude, phase or frequency modulation. The nodes would have to be equipped with modulators/demodulators or modems for short. This allows for increased transmission speed and distance. As each modulated signal has a specific bandwidth almost always considerably less than that of the medium, several signals can be sent at different frequencies. This makes a more efficient use of the cable. A typical use for the extra bandwidth is to include other forms of information such as TV, videoconferencing, etc...

In the Star topology, access control is very easy as only one connection exists between each node and the central node. In other topologies where the medium is shared between all the nodes of the networks, there must exist a way to control transmission of data over the medium. The first method that comes to mind would be circuit switching, where a dedicated link is established between the sender and the receiver. This means that the whole network is dedicated to exchanging information between them. This method is primarily used in the public telephone network. In data communication, things are quite different. When computers communicate together, there is usually a period of activity, followed by a period of inactivity (one computer sends information and waits for the other to answer). During this inactivity, the medium is being wasted as nothing goes through.

A special technique for data communications has been developed called packet switching. In this technique, the is put in a packet and sent over the network. This way, each node has access to the network only when it is transmitting. A control method had to be devised to decide which nodes can transmit in a fair manner. These are divided into two methods, contention and noncontention.

The exact format of the packets varies widely. There are however common elements which make up a typical packet (Figure).

Start of Packet Indicator : This field informs the other nodes on the medium that a packet is being sent, it usually consists of a sequence that never occurs during normal data transmission. It is also used in some systems as a synchronizing signal.

Addressing Information : Each node on the network has a unique address. The address of the receiving node must be placed in this field. Also the address of the sending node is included so that a node can identify the sender. This can be used by the receiver for acknowledgements and/or by the sender to remove a packet that it had already sent. There are also some general addresses for special purposes, like broadcasting to make all nodes receive the packet being transmitted.

Control Information : This field states the purpose of the packet. Special packet may be sent to test the network or inquire about the state of nodes.

Data Field : This is where the information being sent is placed. This field can be of fixed length or variable length.

Error Check : Although the error rates in LANs are very low, an error can still occur. This field allows for error checking. The error-checking scheme may be as simple as a parity bit or more complex as the cyclic redundancy check (CRC).

The first method for sharing the medium is contention. As its name suggests, a node will wait for an opportunity to transmit. When the network is idle, it will begin transmitting. This may lead to several nodes transmitting at the same time, the access method has to offer a way to recover from this situation.

This is one of the earliest access methods, it is known as the Aloha system and it was first implemented at the University of Hawaii. It allowed users spread over four islands to access the central computing facility on the main island. It provided two radio channels, one for terminal-to-host traffic and the other for host-to-terminal

When the host sends information, there is no problem, because it is the only node which transmits, while all other nodes "listen" and receive the information if their address matches.

When a terminal wants to send, it would simply send the packet over the other radio channel. A collision occurs when two terminals send packets at the same time. This collision destroys both packets. Therefore, the terminal waits for an acknowledgment from the host. If it does not receive it within a specific time, it concludes that the packet was destroyed, waits for a random time and sends another packet.

This method is an improvement to the above technique. The sender senses the carrier before it begins transmission. In this case, the odds for a collision are reduced as it can only occur if two nodes begin transmission at exactly the same time. If a collision occurs, the behavior is the same as described above.

Yet another improvement is introduced by enabling the node to continue listening to the network while it transmits. If a collision occurs, all transmitting nodes detect a garbled signal and deduce that a collision has indeed occurred. To make sure that all nodes detect the collision, the nodes detecting it will send a number of random data (called a jam) which lasts long enough to propagate through the whole network. The duration of this jam (called end-to-end propagation delay) should be carefully calculated to avoid wasting bandwidth. When a collision is sensed, all nodes stop transmitting and try again after a suitable retransmission delay. This delay is computed individually for each node using an algorithm called binary exponential backoff. The algorithm is designed specifically to minimize the amount of collisions. This time has to be quantized in steps as large as the propagation delay. The IEEE has standardized a value of 51.2 microseconds for a 10 mbps baseband network.

This method approximates the CSMA/CD method for ring topologies. Each node on the network is equipped with two shift registers. One of them is connected in series with the network. When data is received, it is shifted into this shift register. The address of the packet is decoded and if it matches the current station address it is received, if not, it is shifted out to the next node.

The second register is used when the node wishes to transmit data. First, it waits until the current packet being shifter out is completely transmitted. Second, it checks if there is a sufficient number of empty bits in the shift register before the next incoming packet. If there is enough space, the output is switched to the second shift register containing data to be transmitted, inserting it between two packets. The condition here is that the first shift register must have a many free slots as there are bits to transmit. If the network is lightly loaded, this condition is likely to be fulfilled. If the load rises, there will not be enough free slots and the nodes will have to wait until the load drops again and enough free slots are available.

Both the CSMA/CD and the register insertion allow for very fast transmission times under low load conditions. They also tend to maintain fairness under high loads. Their response times tend to be statistical rather than deterministic.

In this method, access to the network is regulated by making the nodes wait for a permission to transmit.

This method is usually used for ring topologies. It consists of a number of slots circulating around the ring. Each slot is marked as either full or empty. When the network starts up, one station must generate at least one empty slot to circulate around the network. When a node wishes to transmit, it must wait until it receives an empty slot, sets it as full and fills it with its packet. The packet will then circulate through all nodes until it is received. The receiver sets the received indicator and leaves the packet in the network. The packet will then reach the sender which checks the received indicator to know whether the packet has been received or not and sets the slot as empty.

In this system, fairness is ensured by requiring the sender to set the packet as empty after a single transmission, giving opportunity for other nodes to use it once before it can use it again. Under heavy load conditions, where all nodes want to transmit at the same time, each will get a single slot in turn ensuring fairness.

A very important condition here is that the transmission delay through the whole network must be longer than a single packet. As each station receives and retransmits the packet, it causes a delay. If the total delay is shorter than the packet length, the beginning of the packet will return to the sender before it finishes sending it. If needed, a delay buffer may be inserted to avoid this.

A potential problem with this method occurs when a slot is permanently marked as full. The sending could have failed to mark it as empty. In this case, no more transmission can occur. To avoid this state, a single node is given monitor status, which means it monitors the network for such packets. If it does detect one, it sets it as empty to allow other nodes to transmit. Other nodes can be set as passive monitors, ready to take over the active monitor's task in case the active monitor fails.

This method is similar to the one discussed above, but it is a more elaborate one. In this method, a unique sequence called token is passed around the network. Whenever a node wishes to transmit, it "seizes" the token which constitutes permission to transmit a single packet. Afterwards, it must release the token to be passed on to the next nodin sequence.

When used with the ring topology, the token loops through the network as illustrated in Figure. When a loop wishes to transmit, it removes the token and sends its packet. The packet circulates until it arrives to the destination node, which copies it and marks the packet as correctly received. It then sends it to the next node. When the sending node receives its packet back, it must remove it and release a new token. The process is then repeated.

Any topology can be used, with slight adjustments. The ring is the ideal topology, but implementing token passing with the Bus topology is also possible. For a ring, there is only one outgoing connection, so the token does not need addressing information. This is called an implicit token. For a Bus network, the nodes are arranged in a logical ring. Each node is given the address of the next node and whenever it receives a token, it sets the address to this next node. Since this token contains addressing information and is transmitted through the whole network to be received by only one node, it is called an explicit token. The rest of the process continues in a very similar way to the token passing ring.

An advantage found in token passing networks is the possibility for some nodes to have higher priority than others. The higher priority node wishing to transmit can set the priority indicator on a packet to its own level. When the packet returns to the original sending node, it must issue a token at the new priority level while retaining the current priority. Only those nodes on the network with the new priority or higher can seize the token and transmit their packets. After the token circulates around once and returns to the original node, it must seize the higher priority token and transmit a token at the previously retained priority.

A useful application to priority traffic is with synchronous data like speech. Speech transmission requires a packet to be sent at regular intervals. A manager node can regulate this by regularly setting the priority to a defined level allowing the synchronous traffic to be sent on time.

CSMA/CD is very simple and easily implemented on the hardware level. Since each node has to implement the access method, lowering the cost of method implementation can have a serious impact on the cost of the network as a whole. The disadvantage of CSMA/CD lies in the fact that it a contention access method. Priority traffic cannot be implemented, which makes it also unsuitable for real time applications because there is no way to ensure a packet is sent and received at regular time intervals.

Collisions in CSMA/CD have to be handled in a certain matter (previously described) which consumes precious bandwidth. That is why the proportion of time spent handling these collisions must be as small as possible. Since the collision handling time is limited (51.2 microseconds), the packet transmission time should be maximized. It has been shown in practice that under heavy load condition, the efficiency for a packet size of 512 bytes was 97% while the efficiency dropped to 54% with a 4-byte packet. In another perspective, this means that the faster a CSMA/CD network is, the lower its efficiency. This causes a limitation on the improvement of performance by increasing the speed.

I should be noted that these problems only occur over a heavily loaded network. It is normal to engineer CSMA/CD networks to normally operate at a sustained load of 50%. This approach reduces the occurrence of these problems. Care should also be taken when choosing the packet length, if too long a packet is chosen, this results in better efficiency, but longer delays between different packets, which means that larger "chunks" of information arrive separated by large waiting time.

A token passing network does not suffer from collision handling delays by requiring the transmitting node to seize the token before transmitting. However, this does not mean that this method is not without inefficiencies.

Consider the case when only one node wishes to send multiple packets over the network. It seizes the token transmits one packet and releases the token. The token has to circulate around the whole ring (on which no other nodes which to transmit) before the node can begin to transmit again. In other words, the time when the token circulates around the network is wasted bandwidth. As the network is more evenly loaded, less time is wasted. It can be seen that a token passing ring will achieve its maximum efficiency when heavily loaded and that load is spread evenly among the nodes.

Note that this is the exact opposite of CSMA/CD, where best performance occurs when collisions are minimal which occurs under light load concentrated in a single node.

The token passing ring is subject to two related errors. The first is that a node may fail to remove a packet and retransmit a token, rendering the network useless. The second type is when the token is destroyed. This is handled by an additional node (incurring additional costs) called token monitor. This monitor sets a special bit on each packet that goes by. Whenever it receives a packet with this bit set, it means that the packet has already gone through the network at least once. It must be removed and a new token must be issued. This monitor also has a timer which is reset by a packet or a token passing through it. If the timer expires however, this means that the token has been destroyed and a new one must be issued.

As a conclusion, CSMA/CD offers advantages very different from the token passing ring. Before implementing a LAN, careful consideration must be given to how the network will be loaded in order to choose the best media access control method.

Each layer of the OSI protocol stack has a specific purpose and defines a layer of communications among systems. When defining a network process ,such as a file requested from a server, you start at the top where the user makes the request. The request then passes down through the stack and is converted in each layer for transport over the network. Each layer adds sits own tracking information to the packets.

The layers simply define rules that applications, network drivers and network hardware use to communicate over the network.

A protocol stack defines rules that programmers use to create network applications. At the bottom of the OSI model there are basic rules that define communication over the hardware. A programmer working at this level designs drivers for network interface cards. In the middle of the model there are rules that define how networking hardware and software work together. At the top of the model there are rules that define network applications like e-mail programs. They conform to higher-level networking protocols that define how applications on diverse network workstations communicate with one another.

This layer defines the physical characteristics of the cable system. It also defines infrared communications, fiber optics and RS-232-C cable which connect modems to computers.

The physical layer specifies the following:

Electrical and physical connections, how frames become ready for transmission over the cable, how the network interface card gets access of the cable.

This layer defines the rules for sending and receiving information across the physical connection between two systems. The Data-Link layer controls the packets of data. The sending station has to send a packet again if it is corrupted or not received.

The Data-Link layer is divided into two sublayers. The Media Access Control (MAC) layer that manages the transfer of packets to their destinations. The Logical Link Control (LLC) layer that receives packets from upper layers and turn them over to the MAC layer.

This layer defines protocols for opening a keeping a path between systems on the network. It is concerned with how data is moving. Knowing the address of information the network layer determines the best route for trait to the destination. This is important when the LAN consists of a number of segments. If a packet is addressed to a workstation on the local network, it is sent directly there. But if it is addressed to a network on another segment, the packet is sent to a routing device, which forwards it by using the best path through the routers.

It provides the highest level of control in moving data from one system to another. The transport layer provides error detection and correction. If data is missing from the packet, the transport layer protocol at the receiving end arranges with the transport layer of sending system to have packets re-sent.. This layer ensures that all data is received and in the proper order. A virtual circuit, which is something like a guaranteed phone connection, is established in this layer between two systems. The two systems maintain communications of the own during the data transfer session.

This layer coordinates the exchange of information between systems. The name session comes from the communications session that it establishes and terminates. Coordination is required if one system is slower than another. The session layer adds information to the packets about which communication protocol to use and maintains the session until the transfer of information is complete.

The protocols at the presentation layer are part of the operating system and application the user runs in a workstation. Information is formatted for display or printing in this layer. Codes within the data, such as tabs or special graphics sequences, are interpreted. Data encryption and the handling of other character sets are also handled in this layer.

At the application layer, the network operating system and its applications make themselves available to users. Users make commands to request network services, and these commands are put in packets and sent over the network through the lower protocol layers.

Data flows through the protocol stack and over the media from one system to another.

Data starts at application and presentation layers, where a user works with a network application such as an e-mail package. Requests for services are passed through the presentation layer to the session layer, which begins putting the information into packets and opening a communication session between the two systems. Once a session is established, each layer on one system communicates with the equivalent layer on the other system.

Routines at the transport layer prepare the packets for accurate transmission by adding information that help in error detection and correction. This layer provides an interface between the application-level software and the networking hardware. Now a transport layer protocol is selected, such as TCP (Transmission Control Protocol). The transport layer then sends the packet to the network layer.

Routines at the network layer plan the best route to the destination and add routing information to the packets. An internetwork protocol such as IP (Internet Protocol, the internetwork protocol for TCP) may be selected. The network layer sends the packet to the data-link layer, where it is prepared for transport across the connection.

Now the packets are ready for transfer across the network. The network interface card gains access to the cable by using its built-in media access control method CSMA/CD or token passing or another method and sends the packet as a bit stream across the cable. At the remote system, the process is reversed. This process is repeated as many times as needed to transmit the information that is exchanged between the two systems.

So far we have discussed individual components or parts of LANs. A combination of the past sections can be used to implement the LAN. For example, one can use the Bus topology with token passing access control method on optical fibers. In practice, however some combinations have proven to be more useful than others. We now present some of the most important implementations used today. Only the physical layer and data-link layer will be explained as the other layers may change according to the protocol being used.

This is one of the oldest LANs implemented and is still very popular. The original specification defines Ethernet as a baseband network, with Bus topology and data transmission rates of 10 mbps. The MAC used is CSMA/CD and the medium is shielded coaxial cable. Ethernet has developed however, data transmission rates can go up to 100 mbps and broadband is easily implemented by creating several channels. At present Ethernet has developed in many ways, for example thick coaxial Ethernet (10BASE-5), thin coaxial Ethernet (10BASE-2) and Twisted-Pair Ethernet (10BASE-T). For simplicity, the version described here is the original one.

Physical Layer : The physical connection to the network is made through a transceiver. The transceiver transmits and receives data bits, monitors the network (for Carrier Sense) and detects collisions (Collision Detection). Once a transceiver begins transmission, the signal propagates through the whole network without any retransmission by any other transceiver. The Bus topology can be extended over large distances using point-to-point connections which connect distant segments of the Bus.

• Preamble : (8 bytes) This part is included to allow all nodes to recognize that a packet is being transmitted and to synchronize the sender and the receiver.
• Destination Address : (6 bytes) Contains the address of the destination node.
• Source Address : (6 bytes) Contains the address of the source node. The addresses are made up of 48 bits, the first one is called the multicast bit. If set, this bit indicates that the remaining 47 bits are the address of a group, if not, they represent the address of a single node. If all bits are set to 1, then this is a broadcast packet to be received by all nodes.
• Frame Type : (2 bytes) This field corresponds to the control field in the general packet format. It allows higher level protocols to indicate the nature of the data being transmitted.
• Data :(46-1500 bytes) This is the data transmitted. Note that the size can vary widely, allowing each node to send whatever size of packet it wishes.
• Frame Check Sequence : (4 bytes) This is the error checking field. The method used is cyclic redundancy check (CRC).

It should be noted that the minimum size for a packet is 64 bytes and the maximum is 1518 bytes. If a packet is received outside of this range or with a fractional number of bytes, then a framing error occurred and the data is rejected.

Some system restrictions are present for Ethernet. They are :

• The cable from the transceiver to the controller must not exceed 50 meters.
• A segment can be up to 500 meters in length.
• No more than two repeaters are allowed between two nodes.
• Point-to-point connections cannot exceed 1000 meters.
• The minimum length of coaxial cable between nodes is 2.5 meters and 1500m maximum. Using these numbers, the maximum separation between two nodes is 2800m (1500 coaxial + 1000 point-to-point connection + 300 transceiver cable). Note that the transceiver cable is 50 meters per node and 100 meter per repeater for each of the two repeaters allowed.
• The 47 bits in the address field allow for 140 million million address (1.4 with 14 zeros beside it). While this may seem a lot, it is a must because every single station has a unique address across all Ethernets worldwide. This makes the task of connecting Ethernets easy. The maximum number of addresses on a single Ethernet is restricted to 1024.

This network was launched as IBM's local area networking solution. It consists of a star-ring topology, using token passing ring access control.

Physical Layer : The transmission used is copper twisted pair cable. The choice was made to allow customers to use telephone wiring. This enabled the network to handle 72 nodes. Using Data Grade cable, the network can handle up to 260 nodes. Since a ring is used, different sections can use different media. The method of transmitting is baseband with transfer rate of 4 mbps using Manchester coding. Using the star-ring topology, a number of wiring concentrators exists to which nodes are attached. When a new node is added, it supplies power to an internal relay inside the concentrator which extends the ring to include the new node.

Since the signal gets regenerated at each node on the network, some limitations apply to the distance between the nodes and the concentrators according to the medium used as shown in Table. The figure illustrates the maximum distances for Data Grade Medium. Ad for the Voice Grade Medium there can only be two concentrators connected together with a maximum of 72 nodes.

If one network is not sufficient, several networks can be connected using bridges. Bridges are nodes which have normal connections to two rings and provide logical routing of the packets between them, based on the destination address. Several networks can also be connected using a backbone ring located in the center. The center ring serves to connect all the rings together.

Data-link Layer : The protocols used to access the medium are those specified by the IEEE 802.5 committee. The token passing ring is used as previously explained. The packet format is as follows :

• Starting Delimiter : (1 bytes) It is a unique sequence to allow all nodes to recognize that a packet is passing through.
• Access Control Byte : (1 byte) This byte is used for token management. The first 3 bits are for priority reservation, while the last 3 bits indicate the current priority mode. Bit 4 is the token bit, indicating whether this is a token or a packet. Bit 5 is the monitor count bit, allows the monitor to eliminate packets that circulate around more than once.
• Frame Control Byte : (1 byte) The first two bits indicate what type of data is contained in the packet. The remaining 6 bits have a function depending on the type of data specified by the first two bits.
• Destination and Source Address : (2 or 6 bytes) Depending on the size of the network and whether several rings are connected or not, the size of the address is decided. If 6 bytes are used, the first two are the ring number while the remaining 4 are the ring element address. As with Ethernet, there can be broadcasting addresses consisting of all 1's. The first bit indicates whether this is packet is for an individual node or a group of nodes.
• Data Field : (variable) This field carries the data.
• Frame Check Sequence : (4 byte) This field includes the error checking method to detect whether an error has occurred or not.
• Ending Delimiter : (1 byte) This byte consists of another unique sequence to indicate the end of a packet with the last bit indicating whether an error has been detected or not.
• Frame Status Byte : (1 byte) This byte provides a way for the sending node to know if the packet arrived to destination and if the data was received or not. This is done by the Address Recognized Indicator and the Frame Copied Indicator. Since they are not included in the error checking, they are repeated as a mean of checking.

In this type of network a monitor station must always exist to monitor traffic going around the network and to handle errors. Normal errors are handled like lost token, a permanently circulating packet, multiple tokens circulating, transmission error detection (serves to inform the monitor of a defective segment), network disruption and roll calling (to ensure that if the monitor fails another station will take its place).

The FDDI standard defines a ring-structured network that uses a token-passing form of medium access control . Stations are interconnected using full-duplex, point-to-point fiber-optic physical links although twisted-pair cables are sometimes used. An FDDI operates at a rate of 100 Mbps. A station in the ring can transmit frames only if it possesses a special data unit called a token. All the stations on the FDDI LAN receive the transmissions of all other stations, this is called multi-access of the data link. The frames are repeated all the way around the ring and reach every station where each station interprets the destination MAC address in the frame and copies the frame addresses to it.

In addition to the physical layer and data-link layer, there exists a station management function (SMT)

Physical Layer : The FDDI physical layer is divided into a Physical Layer Protocol (PHY) sublayer and a Physical Layer Medium Dependent (PMD) sublayer.

The station in the FDDI terminology is an addressable network component that is capable of generating and receiving frames. Each instance of a PHY sublayer entity and a PMD sublayer entity within a station is called a port. A station can have one or more ports, each port is attached to the transmission medium through a Medium Interface Connector (MIC).

Data-link Layer : It is divided into a Logical Link Control (LLC) sublayer and a Medium Access Control (MAC) sublayer. The packet format is the same as the token ring network.

The FDDI standards permits the use of the IEEE/ISO Logic Link Control standard for this sublayer.

The MAC sublayer is concerned with the protocol used to handle the transmission of tokens and data frames around the logical ring. The difference between the token ring network and FDDI is that a node seizing the token transmits as many packets as desired (called frame) until a specific time limit, the token is then released, allowing other nodes to benefit from the network as shown in figure.

Station Management : Each station has a single station management component responsible for monitoring the operation of the station and for controlling the various station components. The SMT component does the following functions:

• Initializes, inserts and removes stations.
• Manages a station's configuration, its attachment to the FDDI transmission medium and its connection to other stations.
• Isolates and recovers from faults.

So far we have discussed the means by which the LAN transfers data around. This does not mean that the potential is fully exploited. The ability to exchange information has been established with the LAN, the task remaining is to ensure this exchange is used in a constructive way. If we take the example of a simple file transfer, the problems faced will be clearer.

The file is usually too large to fit in a single packet. This means it will have to be broken up and then reassembled. This means additional information has to be sent like file name, size, start of file and end of file. The receiver must be able to distinguish between the data and the control information.

When a packet is received, some time is needed to process it. If the processing is slower than the packet arrival rate, too many packets will flood the receiver and it will not be able to continue receiving. Also, because no network is completely reliable, some information to notify nodes of successful transmission has to be exchanged.

On a single LAN there can be several paths between two nodes. An algorithm has to be devised to pick the best route between the two nodes. There are still more problems like transferring files of different character codes or byte lengths, transferring files between nodes on different networks, the interface of each user may be different, one of the nodes involved "crashed" during the transfer.

This list is by no means complete. We can see that establishing the LAN still needs an extra step to be complete. This step is communication protocols. Many different protocols wereinvented. Recently the OSI model restricted only those layered protocols to be used to insure compatibility and interoperability. The layered approach has proved to be the most successful because each layer communicates with its corresponding layer without worrying about layers above or below itself (as was previously described). Accordingly the IEEE made the 802 project which was further divided into committees each with a certain task on hand. The result is the following :

• 802.1 : This committee produced an overview of the work of the project and defined the LAN reference model. They also dealt with such issues as addressing formats, network management and internetworking.
• 802.2 : This described the logical link control services and primitives to be used in all IEEE-specified LANs
• 802.3 : Medium Access Control and physical layers for a CSMA/CD-based bus network.
• 802.4 : Same as 802.3 but for a token-passing bus.
• 802.5 : Same as 802.3 but for a baseband token-passing ring.
• 802.6 : Specifications for a Metropolitan Area Network.
• 802.7 : This broadband Technical Advisory Group (TAG) advised the other working groups on issues relating to broadband transmission.
• 802.8 : The fiber optics TAG explored ways in which the developing technology of fiber optics could contribute to the work of the other groups.

As there is not enough space to discuss all the different protocols here, only one of the most important protocols will be discussed. This is the TCP/IP protocol. It is one of the fastest growing protocols in the world today and is expected to be the dominant protocol around the turn of the century.

The protocol used at the network interface layer in the hierarchy is known as the Internet Protocol (IP). It provides a facility for the delivery of datagrams to nodes located anywhere on the Internet. This layer has three main areas of responsibility. Firstly, it provides a uniform interface to higher layers, irrespective of the underlying network in use. Because there may be many different networks in the path between source and destination, the layer must also perform addressing/routing, and, since the networks may have different packet sizes, fragmentation/re-assembly of datagrams.

The Internet protocol adds header information to the data presented to it by higher layers. A typical IP header is shown in the figure. The source and destination address fields are 32 bits in length. The IP module inspects these and determines whether the datagram is addressed to a node on this, or another (connected) network. If so, it must translate it (using either static tables, or by a call to a network name server) to an address suitable for the underlying network. If it is destined for a node on some other network, it must select an appropriate gateway and send it there instead.

If the network involved in the transfer is incapable of handling the complete datagram in one chunk, it must be broken up, and sent as fragments. A series of fragments of a single datagram can be identified by the unique combination of the source/destination address, protocol and identification fields. The fragment offset, total length and flags assist in breaking it up before transmission and reassembling it on arrival.

The basic services provided by IP are further extended by the TCP. TCP augments the addressing scheme with port/socket numbers. It also maintains such state information as is necessary to maintain a bi-directional virtual circuit between two sockets on the Internet. TCP transforms the underlying IP datagram facility into a reliable transport mechanism with messages guaranteed to be delivered in the correct order.

The format of a typical TCP header is shown in figure.

This begins with the source and destination ports. The following two fields i.e. sequence and acknowledgement numbers are used for reliability. Each byte in the data stream being transmitted is assigned a number. When a packet is sent, the sequence number refers to the first byte in the packet. The receiver acknowledges, giving the next sequence number he expects to receive. This means that all bytes up to that number have been received correctly and provide means of both ensuring that data is not lost in transmission, and that it is put together in the correct order.

When acknowledging receipt of a packet, the receiver can use the window field to inform the receiver as to how many more bytes can be accepted over and above the one being acknowledged. This prevents the sender transmitting at too high a rate. A checksum field is provided to detect transmission errors. The remaining fields are used in various ways to control the connection.

The consumer of this reliable transport service are applications level protocols such as electronic mail and virtual terminal services.

The Telnet protocol makes use of the reliable virtual circuit facility provided by TCP to provide a network virtual terminal (NVY) capability. An NVT is an imaginary device that consists of a keyboard that can generate characters, and a printer of unspecified width that understands a limited number of control codes (e.g. tab, ring bell, etc.).

A Telnet connection consists of full-duplex conversation between two of these NVTs. The character passed between the two parties use the ASCII character encoding scheme and one of these (character 255) is interpreted to mean 'interpret as command'. When this character is encountered, the following characters are treated as a command.

The user of the Telnet service is responsible for translating the behavior specified by the virtual terminal into actions that are appropriate to the local system. This protocol is often used as a means of remote login to other hosts on the Internet, and is also used by the file transfer protocol (FTP).

One very common use of networks (local or otherwise) is that of transferring files between systems. One of the protocols used for this purpose is FTP. This works by establishing a Telnet connection between the user and the FTP server. A specified set of commands is available to carry out various connection management (e.g. login, accounting, interrupt), context management (e.g. change directory, specify file structure) and file transfer tasks. The data to be transferred travels over a separate Telnet connection and actions on this link are controlled by the server process.

Hosts implementing these protocols normally set up servers at well-known port addresses.

In addition to straightforward file transfer, the FTP protocol also includes commands to direct the data transferred to a particular user as electronic mail. This service is not needed, however, when a dedicated electronic mail protocol is used.

This application level protocol provides reliable mail transfer to/from any host on the Internet. It also has the added advantage that it is independent of the transmission subsystem involved in moving the data. For this reason, it can be used on a variety of interconnected networking environments. In a TCP/IP environment, SMTP makes use of the services provided by TCP. This network could be gatewayed to other networks using, for example, the ISO Transport or CCITT X.25 protocols.

The SMTP layer in one node establishes a connection with another SMTP layer located elsewhere on the network. A simple set of commands is available to identify the sender, specify the recipient(s), and transfer the message body. The receiving node may not be able to deliver the mail directly to a local user, or it may forward it yet to another node on the route to its ultimate destination.

Local Area Networks and their applications by Tangney O’Mahony,

Published 1988 by Prentice Hall

Local Area Networks by James Martin,

Published 1994 by Prentice Hall

Novell Netware 4 by Tom Sheldon,

Published 1993 by Mc Graw Hill