DOWN a tunnel of living flesh, a surgeon鈥檚 laparoscope steers towards a suspected tumour. Minimally invasive therapy, popularly known as 鈥渒eyhole surgery鈥�, cuts costs and reduces trauma. But learning how to use it is difficult, especially in Britain where surgeons, whether they are training or experienced, need a licence before they can rehearse on live animals. This is why Leeds General Infirmary has pioneered another approach.
The hospital鈥檚 Institute for Minimally Invasive Therapy has invested in a data network that provides medical students with ringside seats for operations involving keyhole surgery. The network pipes video images from the tip of surgical instruments, together with the surgeon鈥檚 commentary on his procedure and computer data from other medical apparatus, to lecture rooms up to 2.5 kilometres away from the operating theatre. The technology is called asynchronous transfer mode (ATM), and it is the first practical way to mix video, telephone calls and computer data on the same cable.
Enthusiasts say that ATM, which is still very much in its infancy even after nearly ten years of research, will make the idea of an information superhighway a reality. But others regard it as a distraction that could hinder development of the highway. 鈥淚n Europe we have a unique chance to create the electronic superhighway,鈥� says Jose Breval, managing director for innovation at Europe鈥檚 largest software company Cap Gemini Sogeti in Paris. 鈥淏ut to do this we have to move immediately, immediately. Of course, when ATM comes along we will use it. But we don鈥檛 need ATM for the superhighway today.鈥�
Advertisement
Most major telecommunications companies, such as AT&T and BT, have similarly mixed feelings. Although committed to ATM in the long run, they are reluctant to upgrade their existing networks until they are sure that ATM works 鈥� and that customers will pay for it. This leaves the people running local and specialised networks to do the testing on the ground 鈥� effectively on behalf of the telecommunications giants.
Apart from Leeds General Infirmary, other users include the Department of Defense in Washington DC and the UK Education and Research Networking Association, which is linking universities and research laboratories through a network called SuperJanet. Later this year, Telstra, Australia鈥檚 biggest telecommunications firm, is expected to join the club with the launch of its Experimental Broadband Network to link universities in four cities on the east coast. With AT&T of the US and KDD of Japan, Telstra is also due to establish another experimental ATM network that will link the three countries.
In telecommunications, however, technical standards are everything. Anyone who invents a better technology without persuading others to adopt it merely ends up in what engineers call the Ghostbusters predicament: 鈥淲ho ya gonna call?鈥�
ATM emerged at a meeting in 1988 of the international standards organisation, Comit茅 Consultatif International de T茅l茅graphique et T茅l茅phonique (CCITT), which is now known as the International Telecommunications Union. The telecoms industry had known for some time that such a technology would be needed to exploit the enormous capacity of optical fibre to carry digital information, but its development has been dogged by compromise and confusion ever since.
In theory, encoding voice, computer data or video into streams of digital bits is the great leveller. A bit is the same whether it is part of a telephone call, an airline reservation or a TV programme. But when millions of bits come together from different sources, the flow of traffic makes different demands on the system. In general, video images and telephone calls need a constant stream of bits if they are to appear natural, whereas computer communication such as an e-mail message can quite comfortably be sent in bursts. Because of this, the computer industry developed different techniques for sending information from those of the telephone industry. The hope is that ATM will unite the techniques (see 鈥淭hree into one will go鈥�).
Stripped to its bones, ATM is a way of transmitting data in small packages, called cells, 53 bytes long. Of these 53 bytes, 48 are the data that needs to be transmitted, and the remaining 5 bytes, known as the header, identify the message and tell the network鈥檚 control circuits where it should go.
The cell鈥檚 size was a compromise agreed at the 1988 meeting; American members of the CCITT wanted to squeeze 64 bytes of data into a package, while their European counterparts preferred 32 bytes. As Christopher Cooper of the Advanced Communications Unit at Rutherford Appleton Laboratory puts it: 鈥淔ifty-three bytes is not a number that a computer person would immediately come up with.鈥�
When the technology first emerged, everyone expected public telephone companies to pioneer ATM in their high-capacity Integrated Services Digital Networks (ISDN). But national telephone networks take decades to establish 鈥� and decades to change technological gear. 鈥淚n theory, all the major telecommunications companies have been committed to ATM since the mid-1980s, but it was only intended as a core system,鈥� says Henrik Kjellin. He is managing director of K-Net, a British manufacturer of ATM equipment, which helped to develop the ATM network at Leeds General Infirmary when the hospital could not find an off-the-shelf system to meet its needs.
Britain鈥檚 biggest telecommunications firm, British Telecom will not say when it plans to offer the service throughout its network, and yet it was one of the pioneers of ISDN and a key force in setting the standards for ATM. One reason for the delay, says BT, is that the technology has not yet proved that it is robust enough. In the meantime, like other European telecommunications companies, BT offers a packet-switching service to customers who want to exchange computer data. But the service, known as Switched Multimegabit Data Service, runs more slowly than ATM and is not suitable for voice calls. The same sort of thing is happening in the US, where telecommunications companies are promoting yet another service for computer customers. The service is based on a standard known as Frame Relay, which is a descendant of a widely used standard for data networking, X.25.
Not surprisingly, companies who seven years ago were inspired to develop ATM equipment are now having a hard time trying to promote and sell their products. In October last year, for example, Chase Manhattan Bank postponed its plans to install ATM networks in its international dealing rooms, saying the technology was not ready.
Lee McLoughlin, a researcher specialising in networks at IC Park, a spin-off from Imperial College, London, says the drawback is reliability: 鈥淚f you want [your project] to work in five years鈥� time, then ATM is the solution. Now, no.鈥� But Kjellin sees the bank鈥檚 decision in a different light. He blames the suppliers of other types of networks, such as X.25 and Frame Relay, for trying to undermine ATM technology. 鈥淢anufacturers are protecting their interests,鈥� he says.
Reservations about the technology are not universal, however, and its proponents remain optimistic. In the long term, they hope the technology will underpin a mass market by providing services such as video-on-demand. In the meantime, it must rely on the demands of specialist networks. There is no shortage of such ideas, usually involving metropolitan area networks.
In Edinburgh, a consortium representing the city鈥檚 three universities is about to place a multimillion pound contract for an ATM system. Eventually, the network is expected to extend to schools, libraries, museums, hospitals and private companies, and to provide computer-based teaching and video-conferencing. The network will allow the universities to cut costs by sharing computing facilities, and to gain access to supercomputers in distant cities. The network could even provide a local telephone system.
What the telephone companies and regulatory authorities will make of that remains to be seen: the political problems of breaking down walls between telephone and data communications might turn out to be even more difficult to resolve than the technical ones.
Three into one will go
WHEN optical fibres began to replace copper cables in the 1980s, telephone companies claimed that they had solved the problem of distributing multi-media services wherever they were needed. This was an oversimplification. Exploiting the huge carrying capacity of optical fibres efficiently needed a new generation of research into systems for allocating bandwidth and managing the data traffic.
The easiest way to handle data traffic is to connect a dedicated cable between each caller, a system known as circuit switching. Old-fashioned telephone exchanges worked this way, setting up a temporary physical link of fixed capacity, regardless of what was travelling across it. This was fine when a circuit had the capacity to carry only one telephone conversation. But with modern technologies, circuit switching is cumbersome and wasteful.
For a start, a telephone conversation that sounds continuous to each participant actually consists of about 60 per cent silence. When computers are doing the talking, gaps between messages can get much longer than the messages themselves. To the telephone companies this is wasted capacity 鈥� capacity that they would like to sell.
For the past 25 years, the computer industry has been developing an alternative to circuit switching that minimises the spare capacity, called packet switching. This means transmitting data in blocks, or packets, that may be interleaved with packets from other parts of the network on their way to different destinations. The industry has developed standards to ensure that computers can connect to these networks and recognise the packets of data on them. Different standards apply, depending on whether the networks are local, usually within a building or organisation, or connect a wide area, stretching between cities or countries.
Two technical standards, Ethernet and token ring, dominate the world of local networks. They run at 10 megabits and 16 megabits per second respectively. On wide area networks, the best known packet-switching standard for transmitting data is called X.25, which runs at 64 kilobits per second.
The problem with packet-switching is that it introduces 鈥渏itter鈥� when carrying continuously changing messages such as telephone calls or video signals. ATM, however, represents a merger of the two traditions. It is a way of putting data into packets that behave, when you want them to, like a constant stream. ATM networks transmit data in small packages, called cells, 53 bytes long. The goal of the compromise was speed.
In theory, today鈥檚 optical fibres are capable of carrying data at the rate of up to 2 gigabits per second, compared with copper cable鈥檚 14.4 kilobits. The limit is the speed at which a network鈥檚 control circuits can switch and match packets of data. In an X.25 network, packets are of variable lengths and include protocols to check for errors in the message 鈥� a hangover from the days of copper cables, which are more prone to interference than optical fibres. ATM packages, stripped of this baggage, can blaze through a network鈥檚 switching circuits much more quickly, at up to 155 megbits per second.
ATM standards are continuously evolving. Since 1991, the main force driving the technology has been an ad hoc group of mainly American computer and telecommunications companies, the ATM Forum. It is hard at work agreeing and disseminating the rules for ATM, even before the hardware is ready. A recent breakthrough was an agreement to set a slower standard, 25 megabits per second, for connection to personal computers. This should help to bring down the cost of an ATM connection.
Much work still needs to be done in other areas, however. 鈥淧rinciples associated with congestion control and traffic control are still under development,鈥� says Christopher Cooper of the Advanced Communications Unit at Rutherford Appleton Laboratory. 鈥淭here still remain some issues about what are the best ways to do it.鈥�
One hot area of research is traffic-shaping. Rather as airlines make sure their aircraft are full by selling more tickets than they have seats, telecommunications companies allocate too much traffic to packet-switched circuits. To get away with overfilling both planes and circuits, operators need statistical methods to predict the number of 鈥渘o-shows鈥�. Models exist to predict the statistical behaviour of packets of computer data, but it is more difficult to predict what happens when traffic contains streams coming from many sources.
A related area of research is how to cope with the flow of data overloads circuits in switches and other pieces of hardware. Most calls contain information of different levels of importance: for example, a few pixels in a video screen might be less vital than control information. This distinction is important if parts of the network become congested, for example, when large amounts of data arrive simultaneously at a switch. Researchers want to tackle this congestion by putting information into each cell to say how important it is and whether it should have priority in an emergency.