Peter Cochrane: Optimising to extinction
Economic incentives create dangerous levels of brittleness and a lack of adaptability in the face of rapid technological and market change
We are the only species on this planet with an almost singular focus on optimising everything on the basis of cost, performance and return on investment.
Mother Nature, on the other hand, has learned from billions of years of evolution that optimisation is ultimately destructive to the point of extinction. It fundamentally limits diversity and the ability to adapt and survive in the long term.
Human experience actually mirrors these lessons, but our economic systems encourage, and often demand, slimmed down and tightly focused companies and product ranges. This sees operational, Capex, Opex and RoI targets creating dangerous levels of brittleness are manifest in a lack of adaptability in the face of rapid technological and market change.
Sadly, the relationship between efficiency, reliability and resilience appear to have slipped out of the education curriculum. However, an obvious and easily understood example is an F1 racing car versus a family saloon. The former is expensive, very efficient, but unreliable and not very resilient. On the other hand, the latter is low cost, very inefficient, but reliable and resilient!
An expansion of this basic concept sees us expert in creating highly reliable products by the billions, and all with low-cost components and fit-for-purpose performance at affordable prices. Mobile devices, PCs, networks and server farms are prime examples.
In today's highly competitive global economy, it is important for companies to maintain a broad range of capabilities to ensure they can adapt to technology-driven change. The problem is, this costs money and detracts from the bottom line. Such a focus can therefore see adverse market and shareholder pressure being applied to directly influence the chairman and board members.
But all RoI-focused companies with short-term horizons automatically face even higher risks. The IT and networking industries are a case in point, where energy costs have suddenly become a prime limiter even as there is an increased customer demand for greater resilience and better levels of service. This demands new thinking and radical solutions that challenge many established wisdoms.
For decades the industry mantras have assumed that infrastructure, bandwidth and distance are expensive, but this is fundamentally wrong and can be seen as a throwback to a world of copper.
Optical fibre costs no longer come in at several pounds per metre, they are now a few pence per meter, and so fibre rich networks are viable along with the eradication of powered cabinets in the street. In addition, fibre sharing can also be eradicated as it no longer serves any economic purpose, but limits bandwidth and adds operational risks.
The all-optical passive last mile with equipment only installed at the terminal points has been viable for decades, and reduces energy demand whilst affording infinite bandwidth, extremely low failure rates, and improved resiliency. It turns out that there are similar benefits in the wireless spectrum, as it enjoys less than 20% occupancy. Adopting full-fibre and UWB wireless networking, plus low-energy meshnets for the IoT, sees bandwidth becoming effectively infinite at the lowest energy cost.
At the same time, we have to ask why we continue to code and compress signals into a small amount of bandwidth - it all costs energy, and we can afford to waste bandwidth!
It is as if network engineering mindsets are frozen in some 1950s/60s era. The only way we can effectively reduce energy and operating costs is by simplifying our networks, and getting rid of as much signal processing as possible is just one element. Some 5G towers now consume approximately 10 kW, most of which is for signal processing. And within 10 years we will have 6G, and that development appears to be plodding along the same 5G energy trajectory!
It really is time to think anew.
Peter Cochrane OBE, DSc, University of Hertfordshire