Basic Data Communications Series
Data Networks – LAN Design features
Introduction
Following on from the general introduction to data Networks, we now consider Local Area Networks (LAN’s) in more detail.
The first topic we need to cover is the topology and wiring considerations of a LAN since this is very important first consideration in the practical installation of a system within a building. In particular we need to take into account three main factors:
1.1 Cost
Whatever medium is chosen for a LAN it has to be physically installed and this may be a lengthy and costly process for an existing building. Ideally of course should be carried out before building occupied, but if this isn’t possible then minimum disturbance should be aimed for and this may effect the distances involved and topology adopted.
1.2 Flexibility
The topology should be flexible and easily reconfigurable with minimal disturbance. New nodes should be added or removed.
1.3 Reliability
Failure in a LAN can take two forms. First an individual node may fail – this is not as serious as the second were the complete network fails. In the second case if the individual nodes may function any software making use of the network will be rendered useless. The topology chosen can help by allowing the location of the fault to be detected and to provide means of isolation.
The three major topologies used are star or radial; bus; ring or loop. There are also a number of hybrid network topologies, which combine these features. We consider this next in detail.
2.0 Star or Radial
This consists of a central node to which all others are connected.
It is commonly used in data processing or voice communications networks. A common example is the IBM 370 installation in which multiple 3270 terminals are connected to a host computer or terminal controller via a single length of co-axial cable.
Another example is the office private branch exchange – PBX were each telephone is connected a central PBX by a single twisted pair cable.
In many cases when a building is wired with a star, the cables radiate out to intermediate points called wiring closets. This allows sufficient connection points for a sub-area.
Strictly speaking the two examples mentioned do not qualify as a LAN because they uses a central controller, but are worth mentioning because of the prevalence in traditional networks and influence on star-ring topology.
2.1 Advantages of star
Ease of service
The star topology has a number of concentration points giving easy access at the wiring closets
One device per connection
Connection points are inherently prone to failure. In the star topology, failure of a single connection typically just entails removing a single node from a fully functioning network
Centralized control/ diagnosis
Faults are easily detected at the central controller, which are then easily isolated.
Simple access protocol
Any given connection in a star involves just the node and central controller – no contention problems. Protocols simple.
2.2 Disadvantages
Long cable length
Each node connected to the center therefore large cables and long runs. Congestion in cable ducts etc.
Difficult to expand
Addition of a new node involves a connection all the way to the central node.
Central Node dependency
If the central node fails the entire network fails.
3.0 The bus
The bus consists of a single length of transmission medium (normally coax) onto which the various nodes are attached. This topology is used in traditional data communications networks were the host is at one end of the line and communicates with several terminals. It is also known as a muli-drop line and is the topology used in Ethernet.
3.1 Advantages
Short cable length and simple wiring layout
Because there is a single common data path connecting all nodes. The bus topology allows a very short cable length to be used. This decreases the installation cost and also leads to a simple easy to maintain wiring layout.
Resilient architecture
The bus has an inherent simplicity that makes it reliable from a hardware point of view.
Easy to expand
Additional nodes can be connected to existing bus network at any point along its length. Adding extra segments connected via a repeater can achieve more extensive additions.
3.2 Disadvantages
Fault diagnosis is difficult
Although simple architecture with little to go wrong fault detection not simple. In most Bus LANs – no central bus control – fault finding has do be done along the bus at many points.
Fault isolation difficult
One the fault is found, if a node fault then can be removed but if the fault is in the network medium then an entire section may have to be isolated
Repeater configuration
If backbone extended using repeaters – reconfiguration of existing network may be necessary adjusting cable length, terminators etc.
Node must be intelligent
4.0 The Ring
In the ring each node is connected to only two other nodes ie its physiscal neibours. Data is accepted from one neighbour and passed to the other and data travels in one direction only from node to node around the loop. After passing once around the ring it return to the sending node which removes it. Note that data passes throgh rather than past each node. The signal may therefore be amplified or repeated as it passes throgh. The recipient can easily mark the message as read -–this serves as an acknowlegement to the sender when received back.
4.1 Advantages
Short cable length
Amount of cable is comparivble tot bus and is small relative to a star
No wiring closet space
Only one cable
Suitable for fibre optics
High speed transmission possible. Because data passes in one direction – can use fibre or mixed fibre/cable.
4.2 Disadvantages
Node failure causes network failure
If a node fails then the ring fails
Difficult to diagnose faults
Network configuration difficult
Not possible to shut down a section odf the network when expaning network
Topology effects protocol
5.0 Hybrid Topologies
5.1 The Tree
The tree topology is a variant of the bus. The shape of the network is that of an inverted tree with the central root branching and sub-branching to the extremities of the network. It is normally implemented using co-axial cable as the transmission medium and broadband transmission techniques. The best known is the IBM’s PCN.
The main difference between this type and one made of several bus segments is the presence of the root to the tree. When a node transmits the root (or headend as sometimes called) receives the signal and re-broadcasts it through the entire network. Repeaters are not necessary.
5.1.1 Advantages
Easy to extend
Fault isolation – possible to disconnect whole ‘branches’ of the system
5.5.2 Dis-Advantages
Dependent on root – if this fails entire network down – similar to star in this respect
5.2 The Star-Ring Topology
The star ring are combined to achieve the best of both topologies.
The configuration consists of a number of concentration points connected in a ring
The points in practice consist of wiring closets located on each floor of buiding. From each closet nodes are connected in a star.
Electrically the star ring operates in the same way as a normal ring. The difference is that the physical wiring is arranged as a series of interconnected stars.
5.2.1 Advantages
Fault diagnoses and isolation
Presence of concentration points greatly eases fault diagnosis. Each star can be isolated leaving network fully functioning
Ease of expansion
Modular construction means new sections easily added. Each cncentration point can have extra unused lobes.
Cable-ing
Concentration points connected by single cable. Star sections short cable lengths
5.2.2 Dis-dvantages
Intelligent concentration points required
Cable-ing
The ring cable is critical
6.0 Transmission Media
We now look at the media that is used for LAN’s. We limit the discussion to the types of media used and delay the discussion of the electrical characteristics for the Data Transmission section of the course.
Three main media are used for Local Area Networks: Copper, Fiber and more recently wireless (radio) has been used.
The earliest and most ubiquitous media is of course copper and the most common is in the form of twisted pair cable.
6.1 Twisted Pair
Twisted consists of a pair of insulated conductors wrapped together in a double helix. Each conductor will have identical impedance to ground and therefore it is a balanced medium enabling the use of differential transmission and reception.
A differential signal is one in which two signals are transmitted along each conductor and are subtracted to obtain the information. The advantage of this is that noise pick-up affecting both conductors is effectively cancelled by differencing to zero. There is therefore a high degree of ‘common mode rejection’ by the cable.
As a voice grade medium it is the basis of most internal telephone wiring. For LAN use higher grade twisted pair cable with more stable and accurate electrical characteristics called data grade medium. It is usually shielded with an external foil or braid.
The main advantage of twisted pair is its low cost, simplicity and ease of installation compared to other copper cables. However its data transmission characteristics are not so good with relatively high attenuation and therefore unsuitable for long distance use without the use of repeaters. Also its low bandwidth make it unsuitable for broadband use. It is perfectly adequate for low speed applications were the distances between nodes are small.
6.2 Coaxial Cable
This type of cable consists of a solid center core surrounded by one or more foil or braided wire shields. It is commonly used for TV signals and cable TV systems.
The data transmission characteristics of coax are considerably better than twisted pair with bandwidths of up to 300 MHz possible and so is suitable for broadband use. It also has much lower attenuation and can be used over greater distances. However it is more costly and physically rigid and quite heavy.
It is the basis of the Ethernet transmission media.
6.3 Power-line
A recent development based upon an old idea is to utilize the domestic AC mains wiring as a LAN media. After all the copper exists, why not use it.
In the UK, domestic mains supply is based upon a ‘ring mains’ wiring scheme utilising ‘twin and earth’ cables. These cables have surprisingly good electrical characteristics, better than twisted pairs but not as good as co-ax. The twin conductors are used differentially and can be modeled as an isolated twin line.
In other countries, the position is different. Germany for example uses a ‘tree’ wiring scheme with all three phases of the supply accommodated in the building. Also the wiring uses loose wires contained within a metal tube or conduit – similar to wiring schemes in the UK 50 years ago.
The technology is still immature but initial data rates of 2.5 Mb/s are in development. Much higher data rates are predicted using more sophisticated modulation schemes to overcome the poor channel characteristics caused by severe mis-match problems of the transmission characteristics.
6.4 Optical Fiber
Optical fibers consist of thin strands of glass or similar materials, which are constructed to transmit radiation in one direction from source to destination. The sources are usually light emitting diodes or laser diodes.
Various types of fiber are available and the transmission mode is different, e.g. mono mode, graded index etc.
The bandwidth of the medium is potentially very high certainly well above 100 Mb/s are easily achievable.
The major problems with optical fibers are associated with installation. The fibers are quite fragile and may need special shielding to make them robust enough for an office environment. The connection and splicing of fibers is a difficult process. It is also a relatively expensive technology.
One of the major advantages apart from speed of fibers is there complete immunity to noise hence are a good choice for severe environments, factories etc.
The monodirectional mode of operation make them ideally suited to ring topologies.
6.5 Wireless
Recent developments using short range radio transmission LAN’s e.g. Bluetooth etc. The technology is expensive and there are implications in spectrum usage, EMI etc. The market will prevail and the costs/performance ratios will dominate in the end.
7.0 Media Access Control
7.1 Introduction
We now turn our attention to how individual nodes in the LAN access the media described in the previous section. In the performance section of the course we have noted that there are two main classes of access to a multi-user medium: contention and conflict free. The former treat the media or channel as a free for all whilst in the latter the media is allocated in an orderly sequence.
By far the two most popular and common media access control protocols for Local Area Networks are the token passing ring – which is a conflict free protocol and the CSMA/CD which is a contention protocol.
These two protocols have good performance (refer to curves) and have been the subject of an IEEE standard – IEEE 802. So we will concentrate on these two protocols and look at them in more detail.
7.2 IEEE 802 LAN Standards
In Feb 1980 the IEEE created a standards committee formally known as project 802 to develop standards for Local Area Networks. One of the major objectives of the committee was to work within the scope of the OSI reference model. Accordingly they identified the lower two layers (data link and physical) as those relevant to a LAN’s. They then proceeded to develop a set of functions associated with these layers and their interface to the OSI network layer.
One of the first things to be done was to develop the LAN reference model corresponding to the lower two layers of OSI.
Middle tow layers are OSI Data Link Layer
The OSI Data link layer is divided into two sub layers – the logical link control (LLC) and media access control (MAC). The former is concerned with providing a data-link service to the higher layers while the latter concentrates on providing shared access to the physical layer of the network. The committee then split into 6 groups and two technical advisory groups as follows:
1 802.1 LAN reference model
2 802.2 Logical Link Control
3 802.3 MAC and Physical layers for CSMA/CD based network
4 802.4 MAC and Physical layers for token passing bus based network
5 802.5 MAC and Physical layers for token passing ring
6 802.6 MAN specifications
7 802.7 Technical Advisory Group – broadband transmission
8 802.8 Technical Advisory Group – fibre optics
The CSMA/CD protocol was originally developed by Xerox – their system is called Ethernet and was further developed by DEC, Intel and Xerox. The IEEE 802.3 is virtually identical to this Ethernet system and is ubiquitous for PC based LANs.
The token ring protocol was originally developed by IBM in Zurich and forms the standard for 802.5.
Consider now the basic system description and architecture of the two protocols.
7.3 CSMA/CD – Ethernet
Recall that the basic operation of CSMA/CD is as follows:
Listen for the carrier; transmit only when there is no energy on the medium.
During the transmission listen for a collision; abort the transmission immediately and reschedule if a collision is detected.
Execute consensus reinforcement by generating a jam signal for time tcr.
7.3.1 Physical Layer
The physical layer is responsible for sensing traffic on the channel medium and this is a coaxial cable in the case of Ethernet - thus a carrier sense signal is transmitted to the MAC layer. The physical layer also compares the signal on the medium with the signal that is transmitted and therefore generates a signal to the MAC layer if a collision is detected. These two functions are carried out by the transmit and receive channel access functions.
Data Link Layer
The channel access functions also transmit and receive data bits from the co-axial medium in the form of a Manchester encoded waveform at the correct levels and drive current for the cable. The Manchester encoding and decoding is performed in the transmit data encoding and receive data decoding sections of the physical layer.
A Manchester Encoding scheme is shown below.
1 0 1 1 0 0
A logical true or false is encoded as a positive or negative transition in the center of the bit period. This is also known as phase encoding.
The transmit data encoding and receive data decoding sections are also responsible for generating and removing a 64 bit pattern called a pre-amble that precedes the actual data frame and is used for synchronization purposes
7.3.2 MAC Layer
Logical Link Control
Physical Layer
The framing, addressing and error correction functions are carried out in the MAC layer
Which also acts upon the carrier sense and collision detection signals generated by the physical layer. The link management blocks carry out the collision avoidance using the carrier sense signal and contention resolution using the collision detection signal. The framing, addressing and error detection is performed in the data encapsulation functional blocks.
7.3.3 Packet Format
The data frame or packet in Ethernet consists of a maximum of 1518 bytes as follows.
6 bytes 6 bytes 2 bytes 46 – 1500 bytes 4 bytes
The first 12 bytes give the source and destination address. The 2-byte type field is reserved for the higher; layers and is ignored by the Ethernet protocol. The data field is 46 – 1500 bytes long and the error correction coding is 4 bytes long.
7.3.4 Overall System Architecture
The complete system architecture is shown below.
Transceiver
7.3.5 System Specification
Maximum Bit Rate: 10 Mb/s
Maximum controller to transceiver cable: 50m
Maximum Cable length between two nodes: 1500 m
Minimum Cable length between two nodes: 2.5m
Worse Case Cable propagation delay: 4.33 us/Km
End to end prop delay: 6.5 us
Maximum number of stations: 1024
Worse case round trip delay: 45 us
7.4 Token Passing Ring
The token passing ring protocol is defined in IEEE 803.5 standard and was originally developed by IBM in Zurich.
Recall the basic operation as follows:
A special packet called a token passes from node to node around the ring and in general the receipt of a token represents permission for the receiving station to transmit. If the station has no data to transmit, then the token is simply passed on to the next station in the ring.
If the station receiving the token does have data to transmit then it seizes the token, alters one bit and then appends this to the information it wants to transmit to make up the complete frame. The frame is then passed on to the next station.
The frame circulates the ring until it reaches the destination station, which copies the frame for further processing. The frame then continues on its way around the ring until it reaches the sending station. The sending station can then check if the recipient has seen it. If it has it is removed and a free token is then released onto the ring. This must be done before it can transmit again i.e. it can’t hog the ring.
7.4.1 Physical Layer
The medium used is copper twisted pair, although strictly in the IEEE standard this is not specified. Also in the IBM implementation stations are connected to multi-station access units star fashion, the MSAU’s are then wired together in a ring, whereas the IEEE standard does not specify a topology.
In each MSAU bypass relays may be included to in order to disconnect or unused stations or isolate faulty ones.
The transmission method is baseband and uses differential Manchester Encoding. In differential encoding the bits are encoded as voltage transition as before in the middle of the bit period. However to determine the bit logical value, the polarity in the first half bit time is compared with the second half of the previous bit. If they are the same, then the bit is a logical true, if different then a logical false.
The physical layer then is responsible for transmitting and receiving the bits at the correct levels for the twisted pair medium – which can be of various qualities (shielded etc) and regenerating signals which have been degraded.
7.4.1 MAC Layer
The MAC layer is responsible for implementing the basic protocol, setting up the frames and encoding the data bits. The frame format is shown below.
Token
SD Starting Delimiter 1 byte, marks the start of a frame
AC Access Control 1 bytes
FC Frame Control 1 byte
DA Destination Address 2 or 6 bytes
SA Source Address 2 or 6 bytes
Data Variable length to carry user and ring management data
FCS Frame Check Sequence 4 bytes for error detection
ED End delimiter 1 byte, marks the end of a frame
FS frame status 1 byte
The token consists of the SD, AC and ED fields, these are considered in more detail below.
Start Delimiter SD
Alerts each station of the arrival of a token or data/control frame. The bits are coded so as to violate the differential Manchester encoding scheme to uniquely distinguish the field.
Access Control AC
The first three bits encodes an eight level priority scheme – see later.
The forth bit (T bit) = 1 for a data/control frame = 0 for a token.
The fifth bit (M bit) is the monitor bit which is used to check for continuously circulating packets – see later.
Frame Control FC
Indicates whether the frame contains data or control information. In a control frame indicates the type of control information.
Destination and Source Address SA, DA
Six byte address field
Data
Control data or user data. Variable length, the maximum depends on the ring holding time parameter – i.e. the maximum time a station can hold a token.
Frame Check Sequence FCS
A check sum generated by the source station – depends on content of frame. The receiver recalculates this on frame received and compares to detect errors.
End Delimiter ED
Signals the end of a frame or token. Contains bits to indicate a damaged frame and identify the frame that was last in sequence.
Frame Status FS
Terminates a control or data frame. Includes ‘address recognized’ indication and ‘frame copied’ indication.
7.4.3 Priority System
The eight levels of priority encoded in the AC field allows a user to be designated a high priority station in order to use the network more frequently than lower priority stations.
Only stations with a priority equal to or higher than the priority value contained in the token can grab the token.
After the token is seized, then only stations with a higher priority can reserve the token for the next pass around the network.
When the token is next generated (by last station to transmit) the priority is set to the higher priority of the reserving station.
7.4.4 Fault Management
One station in each ring is chosen as a monitor station. This acts as a centralized active monitoring station and source of timing information for other ring stations.
If a sending device fails after transmitting a frame, then it will fail to remove the frame and the frame will therefore circulate for ever and prevent other stations gaining access, since no free token will be issued. This situation is prevented as follows.
All frames and token issued have the M bit (AC field) set to ‘0’. On passing the monitor station, the monitor re-writes this as a ‘1’ for all frames and tokens with a greater priority than the lowest. All other stations repeat this bit. A token or frame reaching the monitor station with M set to ‘1’ is regarded as garbage and removed by the monitor station.
7.4.5 System Specification
Max bit rate – 4 Mb/s
Max MSAU per ring – 12
Max node to MSAU distance – 300m
Max node to node distance along ring 200m
7.4.6 System Architecture
As mentioned previously, a token ring is implemented as a star shaped ring: the ‘star point’ being located at a concentrator or multi-access Station Unit (MSAU).
In cases were the coverage is insufficient several networks may be connected together by means of ‘bridges’. These nodes have a normal network connection to the two rings and provide logical routing of packets between them based upon destination addresses. Further expansion can be implemented by connecting several ring networks to a high-speed ‘backbone’ ring, which can be implemented in fiber.
7.5 Other LANs
We have covered the CSMA/CD Ethernet and Token passing ring in some detail. Other LANs worth mentioning are as follows.
7.5.1 Cambridge Ring
Work started on developing a LAN at Cambridge University in 1974 and co-incided with the early work on Ethernet.
Consists of an empty slot ring with a transmission rate of up to 10 Mb/s. Uses standard telephone cable media. Widely used for research in UK academic institutions; there have been some commercial implementations of it but not very popular. It is mainly of historical significance.
Data transmission is one way round the ring and sections of the ring are connected using repeaters. Hence different media may be used between repeaters – for example the original had a fiber optic section. The transmission is at baseband and uses four wires or two channels. A ‘1’ bit is represented by a transition on both channels whilst a ‘0’ is a transition on one channel only. Packets are transmitted in time slots.
7.5.2 IBM PC Network
Launched in 1984 for use with IBM PC’s – up to 72 machines. Uses a broadband tree topology. All nodes transmit on one frequency using CSMA/CD. The translator unit at the root of the tree receives the transmission. The transmission is retransmitted on another frequency, which propagates through the net to all nodes.
7.5.3 Wangnet
This is a broadband system using co-ax CATV broadband cable in a simple bus arrangement. The large bandwidth available (400 MHz) is split up into five different bands: Wang Band; Utility Band; PC Service Band; Peripheral Band; Interconnect Band.
7.5.4 Applebus
Uses shielded twisted pair with a bus topology. Transmission speed is 230 kb/s and is limited to 10 machines or so. Uses CSMA/CA access protocol. It is a cheap low performance network intended for Apple microcomputers.
8.0 Local Area Network Utilization
To conclude the section on local area networks we consider the motivation for development of these systems, the advantages of using LANs, Properties of centralized systems and a brief summary of how LANs are used in practice.
8.1 Motivation
The main motivations for the development of Local Area networks can be summarized as follows:
The decreased cost and subsequent widespread use of computers since about 1980 with the introduction of low cost ‘Personal Computers’.
The rapid increase in communications technology – exploitation commercially of low cost techniques no longer restricted to high cost defense systems.
Ease of use and simplification of computers for everyday use. Graphical user interfaces. Computers no longer restricted to ‘boffins’.
The introduction of ‘Information Technology’. Everybody wants information and ‘instantly’.
The thread that links all of these things together is the sharing of resources and information.
8.2 Advantages of LANs
There are two distinct ways of viewing machines connected to a LAN. They can be treated as individual autonomous machines which occasionally exchange information or as a single distributed computer system, which is made, up of a number of nodes or access points.
The following is a list of potential advantages for LANs
Sharing. This is the key attractiveness of LANs. The ability to share resources between users of the network – both hardware and software are available to all users.
Incremental growth. A computing system based around a LAN has the ability to expand and contract easily. New resources can be added as needed or as finance becomes available.
Computing power can be placed were it is needed and can be autonomous. A centralized system however is not autonomous. If it fails then all users fail.
Redundancy. Two copies of the same file can be stored in different locations. Thus if one node is temp unavailable, the information can be obtained elsewhere. Similarly with resources i.e. a printer may be temp removed and output redirected to another on the network.
8.3 Properties of Centralized systems
Some properties of centralized systems that should be preserved for LANs
Sharing Information
Communication. Email etc
Expensive peripherals easily shared
Powerful ‘number crunching’. For CPU intensive applications were a single PC is inadequate can be distributed among the network.
8.4 Problems with LANs
The above list does not come free of charge.
Extra software is required for a distributed system to perform the same functions as a centralized system.
System backup. In a LAN based system automatic tracking of backups is not easily carried out.
Security. Centralized systems have this problem but the increased susceptibility of information to attack while being moved round the network is a problem. Thus encryption is required with extra cost and complexity.
Need for standards have to be imposed. Non-compatibility of different OEM’s equipment.
System failure. Particularly if software is shared across the network.
8.5 Application Areas of LANs
File transfer, data sharing.
Office automation
Industrial Control
Distributed systems