Last Modified: ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?



Homepage / Publications & Opinion / Archive / Articles, Lectures, Preprints & Reprints

If Networks Don'T Live, They Will Die
The Weekly Diamond, Japan, 16 October 1999
British Researchers Plan Future Electronic Society
By Peter Cochrane

It has always seemed to me that a key feature of exponential growth is that it tends to go unnoticed for a very long time, and then suddenly it creeps up and makes a dramatic impact. Perhaps a good analogy is the tidal wave which looks insignificant on the horizon, and by the time we judge it to be dangerous, it is far too late to run away. As a mechanism of change it is obviously so much more dramatic than linear, or even a geometric accumulation, and it is certainly the case with the Internet and computing power that both their arrival and the impact were unexpected. Suddenly the Internet and PCs are all pervasive and we now depend upon their abilities for our existence. It took 100 years for telephone networks to become an established and essential utility, whilst mobile phones achieved that distinction in less than 10 years. However, it is also clear that neither the telephone network nor the Internet are fully satisfactory for the new century. In the former case; the limited connectivity, structural rigidity, lack of responsiveness to change, restricted bandwidth and high costs are limiting to the point of extinction. Whilst our wireless networks are mostly unusable for mobile to mobile phone calls because of their very poor quality. In the case of the Internet it is the inability to control latency, coupled with poor hardware and software reliability, that precludes real time applications involving any form of human life risk.

Our history and experience sees us tending to build very structured and complex systems to complete essentially very simple tasks. In contrast, mother-nature does the reverse. Cellular animals such as slime mould or jelly fish, or communities of ants and bees, complete incredibly complex tasks with little hierarchy and very simple communication and control channels. In our defence, mother-nature has been engineering for nearly 2Bn years and she has learnt of few tricks on the way. This prompts the notion that we might be able to learn from her and even reverse engineer some of her solutions and techniques.

Artificial life is a new area of work that is a considerable step beyond genetic algorithms. It involves sexual reproduction, competition, survival of the fittest, and evolution. At the forefront of this field up to 10,000 parents simultaneously contribute genes in the creation process of progeny. Virtual children are judged relative to the attributes of their parents. If the under perform a parent they are terminated. If they outperform a parent, then the parent is terminated and they contribute to the next generation. As a result, the evolution process in this silicon world is more directed and far faster than our carbon experience. And for stable problem sets it turns out that the optimum number of sexes seems to be four. So whilst an unknown or new class of problem might initially be attacked by 1000s of sexes, a stable solution emerges at around 4 sexes. This might seem a brutal regime, but it is less so than that of the lion and the wildebeest, and turns out to be far more adaptive.

Using these techniques to solve network routing problems saw a 1.6M lines of high level code replaced by less than 1000 lines, and a physical plant reduction of 20% for the same level of customer service. It also saw the fastest solution to the Travelling Salesman Problem, with an extension to 'N Dimensions' more easily accommodated than the classical and very limited mathematical and algorithmic approaches. Moreover, the solutions are dynamic, able to continually adapt to change, and cope with the new activities of people and machines as their habits and technologies become ever more sophisticated. This technology is also ideally suited to multi-level searching, routing, agents, network management, and humanising machine interfaces.

Looking to the future; it is clear that the world will become even more complex and our dependence on networked technology will increase. Artificial life seems to be the only technology able to meet the requirements of such a world. Step on 1, 10,100, 1000 or 10,000 ants and the colony continues to function. But change one character in one line of 1,000,000 lines of a well structured high level software program, and the chances are that the system will collapse or at least seriously malfunction.

Today we have evolved to a state of near engineering lunacy in computer and telephone networks with an unimaginable mix of technologies. IP transported by X25 or frame relay, over ATM, via SDH or PDH, on a new backbone of WDM. This is compounded by network control and management software overhead that dwarfs the cost of the hardware. This is grossly uneconomic and unsustainable on all fronts. For the future to work, networks will have to slim down to IP over WDM with integrated and minimalistic control and management software overhead. Moreover, the software will literally have to be alive to the fast changing demands of the customers, be they people or machines. In the near future, networks that fail to come to life will simply die.

All materials created by Peter Cochrane and presented within this site are copyright ? Peter Cochrane - but this is an open resource - and you are invited to make as many downloads as you wish provided you use in a reputable manner. Website maintained, hosted and designed by USP Networks Events Review