Dial up Modem  Leased Line  ISDN  Kilostream  Megastream  X.25
Wireless Links - Infrared Laser  Microwave  Radio
Frame Delay  SMDS  ATM
Serial Interface Types

• V.24
• V.35
• X.21
• G.703

The cheapest and simplest type of WAN connection is a dialup modem link using POTS (Plain Old Telephone Service). This uses an ordinary modem at each end linked to the networks at each end via a bridge or half router. The modem (MOdulator/DEModulator) is used to convert the networks digital signals to the analogue signals used by a telephone line. With current modem technology, the speed is limited to 33,600 bits per second (V34) multiplied by a theoretical compression ratio of 4:1 (using V42bis data compression) giving a possible speed of 134,400 bits per second. However, this "speed" has an enormous amount of overhead, so a more realistic theoretical throughput maximum is 128,000 [or 128K].

Speeds of up to 115,200 bits per second are far faster than many computer serial ports can cope with. Typically a computer (most Macintoshes for instance) can cope with up to 57,600 bits per second at most (the new Centris 660AV and Quadra 840AV Macintoshes from Apple can cope with up to 64 Kbits per second). Most PCs use a 16540 UART (Universal Asynchronous Receiver/Transmitter) and have even greater problems (especially when running Windows) and can only cope with up to 9,600 bits per second. If the PC has the higher performing 16550 UART serial chip then it can cope with up to 19,200 bits per second. To go even faster, companies like Hayes (famous for defining the defacto standard modem command set) have developed serial cards for PCs which can run at even greater speeds. Hayes' card is called the ESP (Enhanced Serial Port). Many newer computers come with higher speed serial ports, but not always.

In order for the communications equipment to be able to determine when one byte or character (eight bits) of information starts and ends, extra information must be sent over the link. This can be asynchronously (where each eight bits of information is preceded by a start bit and followed by a stop bit which amounts to a 20% overhead), or synchronously (where no start and stop bits are required and instead a special synch character is sent periodically to 'synchronise' the timers at both ends).

A dialup modem link has the exact same charges as an ordinary telephone call (i.e. you are charged for each minute of usage), indeed it uses the same telephone lines as ordinary voice and fax communications.

Some people may say that an ordinary telephone line is only capable of operating at up to 2400 baud. However, baud, despite many people mistakenly thinking so, is not the same thing as bits per second. Baud means the number of state changes in the signal per second. A single baud can, and at higher modem speed does, represent more than one bit of information. Therefore ordinary telephone lines can and do, operate at considerably more than 2400 bits per second.

These same people might also say that in order to get high speeds such as 14,400 or 28,800 bits per second reliably on a dialup telephone line, you must get what is sometimes called a 'conditioned line' from the PTT. These days this is not necessary since most dialup lines are now on digital telephone exchanges which means they are already a higher quality than conditioned lines used to be. Furthermore, modern modems are much more capable of operating even under bad line conditions due to advanced circuitry and error correction protocols such as V42.

The original option for WAN connectivity was the leased line, point-to-point between remote devices, rented at a fixed charge irrespective of line usage and carrying analogue or digital signals.

Originally you had to buy special leased line modems from the PTT. These days most high end desktop modems can be used for both dialup links and leased line links.

A leased line link operates in exactly the same way and at exactly the same speeds as a dialup link (i.e. a modem is required at each end) but instead of being charged for each minute of usage, you are charged a fixed rental per quarter. The only restriction of a leased line is that it is a fixed link between two sites (i.e. you cannot dial anyone in the world).

ISDN is a digital dialup service (unlike an ordinary telephone line which is analogue). Since it is digital you don't need a modem to convert the signals. Instead a device called a Terminal Adapter is used (this is sometimes built into the ISDN device).

Two types of ISDN are available, ISDN2 and Primary Rate ISDN. ISDN consists of two different types of channels, B channels which operate at 64 Kbits per second and are used to transfer data and voice traffic, and a D channel which is used for control signals. ISDN2 has two B channels and one D channel. Primary Rate ISDN originally consisted of 30 B channels, however it is now also possible to obtain it with only 15 B channels; this has enabled smaller companies to be able to afford it. Currently in the UK, both BT and Mercury offer Primary Rate ISDN services, but only BT currently offers an ISDN2 service.

It is possible to do the reverse of multiplexing (splitting one large channel into several smaller ones), and 'aggregate' several B channels together. In the world of Video Conferencing, this is called Inverse Multiplexing, or IMuxing. So with ISDN2, both B channels could be aggregated to form one pseudo 128 Kbits per second channel, and with Primary Rate ISDN up to 30 B channels can be aggregated together, totalling 1,920 Kbits per second (or just under 2 Mbits per second). This is roughly equivalent to a Megastream link.

KiloStream is a 64 Kbits per second digital leased line (some older Kilostream lines run at 9,600 bits per second). It uses a device called an NTU (Network Termination Unit) which you would then connect to your network bridge or half router. The NTU has either an asynchronous V.24 interface (for the older 9,600 or 19,200 bits per second versions) or an X.21 interface for the current 64 Kbits per second lines which can run asynchronously or synchronously.

The name Kilostream derives from the fact that its speed is measured in terms of Kilobits per second. With Kilostream you pay a fixed sum per quarter and can use it as much or as little as you want without incurring extra charges. When deciding whether to rent a leased line like a Kilostream, you need to look at how much time per day you will be using the link. If you will only be using it for a couple of hours a day then a dialup link telephone link or ISDN might be more cost effective. As with a leased line, a Kilostream is a fixed link between two sites (i.e. you cannot choose to one day be talking to one site and then the next to another).

BT have recently started offering a new service called Kilostream 'N', which is N times 64 Kbits per second (e.g. 4 times 64 Kbits per second which is 256 Kbits per second).

Megastream is a 2 Mbits per second digital leased line. While it is certainly possible to use all of this bandwidth as a single WAN channel, it is more common for it to be multiplexed to allow several different simultaneous channels (e.g. Ethernet, serial links, voice links between branch switchboards, etc.). Megastream uses a G.703 interface which has two BNC connectors, one is used for transmitting data, the other for receiving.

The name Megastream derives from the fact that its speed is measured in terms of Megabits per second. Like Kilostream, you pay a fixed sum per quarter and can use it as much or as little as you want without incurring extra charges. Also, as with Kilostream, a Megastream is a fixed link between two sites (i.e. you cannot choose to one day be talking to one site and then the next a different site).

The universal standard associated with the packet switching technique is the CCITT's X.25, dating back to 1976 and revised in 1984. It defines the interface between Data Terminal Equipment, any programmable user device such as a host computer, and Data Circuit terminating Equipment, such as packet switching equipment.

X.25 is an 'access layer' protocol: the CCITT recommendations are for data entering and leaving the network only - they do not include anything happening within the network itself.

Across this access link, multiple simultaneous virtual calls can be established. These logical connections, known as switched virtual circuits, carry the data packets - default size 128 bytes - escorted by the link protocol, which performs any necessary error detection and error correction.

X.25 conforms to the ISO seven layer model, complying with the physical, data link or frame layer and the network or packet layer. Each layer is functionally independent, responsible for one aspect of the total communication, accepting one or more inputs, and producing a single output.

Once data has been transferred, the call is 'cleared', just as a telephone call hangs up, disintegrating the virtual circuit and leaving the participating devices free to establish a connection with any other device in the network. Users are billed for the quantity of data sent - not link time.

Devices in the X.25 environment never try to understand data content; they encapsulate data in packets and add control information in the form of a 'header', so anything from digitised speech to graphics can be sent. It does not handle 'real time' speech, however, due to the queuing system adopted by the switches.

Several standards relate to X.25. X.3, for example, supports packet assembler/disassembler (PADs) which allows multiple devices which are unable to support X.25 - such as ASCII terminals - to attach to an X.25 network.

X.28 governs procedures between the DTE and PAD; X.29 outlines the exchange of control and user data between a packet node DTE and a PAD; X.32 allows dial-up devices to communicate as X.25 DTEs to X.25 networks; X.121 addresses X.25 DTEs and their associated network connections via a 14 digit number.

It is possible for a company to have a private X.25 network using its own lines and switches. Or you can use one of the public X.25 networks (e.g. BT's PSS - Packet Switch Stream, or France Telecom's TransPac service).

X.25 typically runs at speeds of up to 64 Kbits per second but it is possible to run faster links.
 

Wan Link	Speed	Usual Connector
Dial up Modem	28.8 Kbps	V.24
Leased line Modem	28.8 Kbps	V.24
Basic Rate ISDN	2 x 64 Kbps	V.24 or X.21
Primary Rate ISDN	30x 64 Kbps	G.703
Kilostream	9.6 Kbps	V.24
Kilostream	64 Kbps	X.21
Kilostream N	N x 64 Kbps	X.21
X.25 Packet Switching	Varies, usually < 64 Kbps	V.24
Megastream	2 Mbps	G.703

Infrared laser beam links operate in line of site only between two fixed points. The maximum range over which they can operate is one kilometre. Unlike Microwave links, an Infrared laser link can transport LAN traffic at native speeds. As well as a 2 Mbps link, there are option for 4 and 16 Mbps Token Ring, and 10 Mbps Ethernet. With Infrared laser no license is needed to operate it. Typically due to the short range, an Infrared laser link is used in cities to link closely located buildings, or where sites are separated by a road or river.

The Infrared laser link can be affected by weather conditions. While rain does not pose much of a problem, blizzards and severe fog can cause the beam to be interrupted.

The main advantages of microwave compared with traditional ground based links are cost and configuration flexibility, in that order.

A microwave network can be expanded just by replacing modules, whereas a land line requires new links to be provided to increase bandwidth. What's more, the nodes of a microwave network can be shifted more readily just by moving equipment.

In order to operate a microwave link in the UK, a license is needed from the DTI (Department of Trade and Industry) who are responsible for licensing all radio, microwave and telephone systems.

There are three potential hazards, although these can usually be overcome by careful planning. First, there have to be available frequencies; second there has to be line of site between stations (which can be a problem in undulating or urban terrain), third, signals are affected by torrential rain (although not by fog).

The first of these problems arises when there are two different links following a similar line of site, when the frequencies may interfere with each other. Although nodes are usually at most 20 kilometres apart, the signal retains sufficient strength over considerably greater distances to cause interference. In order to avoid this, each microwave supplier is allocated a range of frequencies which they can allocate amongst their users. Microwave links operate in the region of 22 GHz.

A difference of two degrees in direction is sufficient to avoid interference, which means each frequency could in theory be reused 90 times (not 180 times) by going round the compass. This number can in fact be doubled because there are two polarities for each frequency.

The second problem of hills or buildings getting in the way of line of site can be overcome by using a repeater (basically a second pair of microwaves dishes which splits the link into two halves one half on each side of the obstruction). In some cases however this might not be cost effective.

The third potential problem, bad weather, is only a major issue in parts of the world that frequently experience rain of tropical intensity. In more temperate regions such as the UK, potential rain intensity is a factor that has to be considered in balancing reliability against the length of each hop. In a dry climate for example, it might be possible to have nodes 25 kilometres apart. In wetter parts of the UK, such as western Scotland, it might be necessary to come down to 10 kilometres in order to maintain satisfactory reliability of say 99.9%.

Currently radio is only used for links within the same building. This is due to several problems. First, use of the radio spectrum is intensely regulated (in the UK by the DTI) and very few frequencies have been made available for this purpose so far. Second, the speed of radio links is much lower than other types (typically only up to 1 Mbit per second). Third the distance radio links can operate over is quite limited (partly due to the limitations of the frequency being used, and partly to avoid interfering with someone else's signals).

In the UK 2.4 Ghz has been allocated for use by radio based networks and at this frequency, the range is typically only 80 to 90 metres within a building or up to 1 kilometre using a highly directional aerial for line of site links. At 2.4 Ghz, speeds range between 1 and 2 Mbits per second.

A second approach is to use a cellular radio system (like that used for car telephones). As cellular phones use an analogue system a modem is either connected to a cellular telephone or a special cellular modem is used. As any regular user of a car telephone knows, the quality of the link can be very poor. To get round this, Microcom, (a modem manufacturer) developed a standard called MNP10 (Microcom Networking Protocol 10), which helps get round this problem by using error correction and reducing and increasing the speed of the link based on the quality of the signal.

In the near future a new cellular radio system called GSM (Global System for Mobile communications) which is completely digital and therefore more reliable, may help increase usage of cellular radio for linking computer networks.

Frame Relay is touted as the ideal solution for linking local area networks. Often described as a turbo-charged X.25, it transmits at high speeds and is designed to handle the bursty traffic LANs usually generate.

The advantage of Frame Relay is its better use of bandwidth than is seen in other technologies, such as X.25. Frame Relay can send data at the full speed of the line that it runs on. So if it is connected to a 2 Mbits per second line, it can take the full 2 Mbits per second bandwidth.

A Frame Relay backbone network can be implemented via a switch or a multiplexer. The individual vendor offerings tend to be determined by the existing installed base: those with a large base of multiplexors are building Frame Relay cards to fit inside their existing products; those which have concentrated on-packet switching sell Frame Relay switches.

Access to the network is normally achieved through the Frame Relay equivalent of a packet assembler/disassembler (PAD) - dubbed a Frame Relay Access Device or FRAD.

One characteristic of the technology is that it has no in-built error correction and checking procedures so as to enable fast transmission - if it detects an error in a frame, it will simply drop it. Dropped frames should not be too much of a problem since protocols such as TCP/IP, SNA, or AppleTalk should take care of the lost frames and provide a mechanism for resending.

It is generally not worthwhile trying to run Frame Relay over analogue links, in spite of what some vendors say. The problem is that the noise on the line is likely to cause errors and lose data, and the benefits of Frame Relay will be lost, too. For example average errors on a digital line are one in a million, but on an analogue line much higher error rates of one in a thousand are possible.

Frame Relay has an overhead of about eight to 12 Kbits per second of control data. It is therefore only advised to use Frame Relay with links of at least 128 kbits per second. Current implementations range from 64 Kbits per second to 2 Mbits per second.

A characteristic of Frame Relay technology is that it can be set to the speed at which the data would normally travel, yet permit the data to "burst out" of that range and use up more bandwidth - if that extra bandwidth were available. One example would be on a public Frame Relay network where 2 Mbits per second lines might be shared between two or more users. Each user would have a committed information rate of 64 Kbits per second but, if no-one else was using the service at the time, the data could burst its 64 Kbits per second barrier and take up more bandwidth. This is known as fractional service.

The network also needs to set up a priority queuing system, in order to ensure that there is no conflict of interest between the different users. On a public network, this is designed to ensure that customers who have paid for higher bandwidth get priority over others. This means that if a user who is paying for 64 Kbits per second momentarily sends at 1 Mbits per second, then that user does not hold up another who has paid for more bandwidth. Private network users might employ the facility to give preference to certain end-user departments.

In order to make the priority system workable, there are normally two "burst limits" set by the network manager. At the lower burst limit, the network marks all the frames which are eligible for discarding; at the higher burst limit, the user has burst too far and all the data above the limit will be discarded.

On most public Frame Relay networks, tariffs will be determined by the committed information rate and the distance. Normally, it is a flat rate that does not take account of bursting data. This is partly because billing systems that count the number of frames transmitted are only starting to appear. BT Tymnet (part of British Telecom) currently offers a service, marketed as ExpressLane, which is in operation in a number of countries including the UK and US. Other leaders in Frame Relay services are Infonet and Sprint.

Switched Multi-megabit Data Service provides LAN-like features in the wide area, with high speed data interchange services and broadcast style flexibility. But not only is the technology highly suited to LAN interconnect applications, it is also geared to carrying what's known as isynchronous data; time critical, asynchronous voice traffic and interactive video used in applications such as video conferencing.

SMDS technology supports a deterministic service, which means data - including isynchronous data - is guaranteed to arrive at its destination within a minimum time frame. SMDS can multiplex data deterministically because it is cell, rather than packet based: an SMDS cell has a fixed length of 53 octets so the rate of data throughput doesn't fall bellow a certain minimum. If there is competition from other devices to use the available bandwidth, cell relay technology can reserve part of that bandwidth for time critical data, such as voice. It is also fast supporting initial speeds of 45 Mbits per second (T3) in the US, and a likely speed of 34 Mbits per second (E3) in Europe.

The use of cell-based technology enables SMDS to carry diverse types of data, possibly to different destinations, over a common service. Data traffic from a LAN, for example, could be presented as a packet via a Frame Relay or HDLC stream. The interface from the HDLC stream to the SMDS service would fragment the packet into fixed length cell, each with address and sequence information, and transmit the cells over the SMDS network infrastructure. At the destination, the cells would be reassembled into a packet and passed on to the target LAN. Voice traffic would be presented in digital form, fragmented into cells, and transmitted in a similar way.

ATM like SMDS, also uses 53 octet cells. So it will be relatively simple for carriers to migrate from SMDS to ATM. To users it will be nigh-on transparent, except that the response times will be potentially even faster.

BT is currently implementing a SMDS network to link UK Universities called 'Super JANET' as a faster successor to JANET; the Joint Academic NETwork.

Asynchronous Transfer Mode was conceived in the early 1980s as a technology for switched public telephone networks, ATM is now being hailed as the key to interconnecting the world.

This technology, the zealots say, will offer easy, seamless connections between your LAN and the public telephone network. ATM will provide your desktop system with almost unlimited bandwidth on demand, and will enable multimedia applications with real-time, high-fidelity audio and video. ATM supporters are even predicting that the technology will improve the performance of your Ethernet. The truth is that the technology will do all these things, but not until all the pieces are in place.

ATM is a method of breaking up data into 53-byte cells, or packets, and transmitting them from place to place on a network over a series of switches. In operation, it resembles the transporter on the Starship Enterprise: Your data is disassembled at one point, transmitted to the destination, and reassembled in the proper order.

ATM is a universal system: Anything can go in one side and come out the other. You're not limited to one kind of data. The system neither knows or cares what's inside the cells. It can be voice, audio, ASCII text, a series of Ethernet or FDDI (Fibre Distributed Data Interface) frames, or a combination of the above. Each data type has different characteristics (e.g. computer data is bursty, but video is continuous), and ATM technology can accommodate those differences.

ATM supplies bandwidth on demand. You can grab as big or as small a chunk of network bandwidth as you need, and pay for only as much as you use.

Like drivers in a transcontinental road rally, ATM cells start out knowing where and when they must arrive at the end of the journey, but are free to choose from various routes to get to its destination. In fact, different cells from the same chunk of data can take different routes across the network.

Packet-switching technologies, such as X.25 and frame relay, also work this way. This contrasts with circuit-switched systems like the telephone network, in which a circuit is established between two people for the duration of the conversation. With ATM, a cell's route may not be determined when the cell leaves its source, but rather invented dynamically as it makes its way through the fabric of switches that make up a network. This allows data to be dynamically re-routed if there's a network failure.

Traditional network routers must be smart, because they deal with cells of information of varying lengths. ATM switches, on the other hand, deal with cells that are always the same size. This allows ATM technology to perform switching in hardware. Because hardware switching is faster than software routing, ATM switches are universal and blindingly fast. While Ethernet and token-ring LANs run at 10 and 16 Mbits per second, respectively, ATM speeds begin at 155 Mbps and increase rapidly from there. First-generation ATM switches run at 2.4 Gbits per second.

A typical LAN is a shared medium. Instead of relying on a dedicated line between workstations for communications, it uses a single party line to carry each conversation intermixed with dozens of others. Every node in a LAN uses the same line to communicate on the network.

Adding nodes to the typical network increases the load, but not the capacity. Because the entire network is a single line, the aggregate bandwidth (i.e. the total bandwidth of the network) and the peak bandwidth (i.e. the bandwidth of an individual line) are the same.

Even with a fast, fibre-based network like FDDI running at 100 Mbps, you're sharing that throughput with every node in the network. A shared-medium LAN is only as fast as its slowest node.

In contrast, ATM is a switched medium. To understand switching, think about making a telephone call. You dial my number, I answer, we talk. In effect, we have a private dedicated line connecting our two telephones. In fact, our telephones are connected through several switches, which are universal connections. We can be connected to any one of millions of other telephones throughout the world. Even though telephone lines have low bandwidth, the capacity of the huge switched network is astronomical. Like the telephone network, ATM also uses switches. The theoretical aggregate bandwidth of an ATM switch is equal to peak bandwidth of each line multiplied by the number of lines going into the switch (e.g. 10, 155 Mbps lines = 10 x 155 Mbps, or 1.55 Gbps). A modest-size switch can have a startlingly high aggregate bandwidth.

The V.24 signalling scheme is used with serial interfaces such as RS232. It can be used asynchronously or synchronously. It is used for slower connections, e.g. Modem up to 57,600 bits per second, ISDN with a terminal adapter up to 64 Kbits per second, X.25, and slower Kilostream links (9,600 or 19,200 bits per second).

Most micro-computers have V24 - RS232 interfaces. Typically a computer (most Macintoshes for instance) can cope with up to 57,600 bits per second at most (the new Centris 660AV and Quadra 840AV Macintoshes from Apple can cope with up to 64 Kbits per second). Most PCs use a 16540 UART (Universal Asynchronous Receiver/Transmitter) and have even greater problems (especially when running Windows) and can only cope with up to 9,600 bits per second. If the PC has the higher performing 16550 UART serial chip then it can cope with up to 19,200 bits per second. To go even faster, companies like Hayes (famous for defining the defacto standard modem command set) have developed serial cards for PCs which can run at even greater speeds. Hayes' card is called the ESP (Enhanced Serial Port).

Another problem with RS232 is that the standard does not specifically state what type of connector should be used. This sometimes makes connecting devices from different suppliers a very complicated process. For example, Macintoshes use a mini Din 8 pin round connector, PCs use a DB9 or DB25 connector, the old BBC Micro used a DIN6, etc. Most communications equipment fortunately, uses the more common DB25 connector.

Used with medium speed digital connections up to 64 Kbits per second. The type of connector used is known as a MRAC connector.

Used with higher speed connections, e.g. a single Kilostream at 64 Kbits per second or Kilostream 'N' at n x 64 Kbits per second, or dual ISDN2 channels (128 kbits per second).

It uses a DB15 connector.

Used with very high speed connections, e.g. Megastream, or Primary Rate ISDN.

It uses twin BNC connectors, one for transmitting and one for receiving.

G.703 Connector is a pair of BNC connectors.