Ian Bitterlin, visiting professor at Leeds University, former Emerson CTO and data centre evangelist, offers his take on the some of the most important and unanswered questions about the edge, and explains why hotels are the perfect sites for powering the new edge ecosystem
Who is going to own the edge? I would bet on the telecom providers, both fixed-line and cellular mobile. Who else is better placed, or better experienced, in rolling out such a distributed comms network? They have done it before, originally installing 20,000 cell sites (in the UK alone) for cellular-mobile and, more recently, look at broadband cabinets in street-furniture – which must be the perfect definition of edge.
They own the next layer in the infrastructure – the telephone exchanges, and those are already connected with bulk-fibre. They already have the man-power with all the skills needed, vans, call-out system etc., and, as 5G arrives the edge will ride-in on its coat-tails.
Losing sight of the edge
But who owns the edge isn’t the most interesting question. From my perspective I feel lots of ‘heat’ from the marketeers telling us all about their edge products and their readiness but see very little light. What I do read and hear is vendors telling us what they hope will be the edge, based on what maintains their existing business model and product lines.
No, the question should first be, what will the edge look like? I certainly don’t think it will look like a mini data centre – unless you define a data centre as one ICT cabinet housing compute, temporary/buffer storage, and I/O. The cabinet might not even be housed in a room. It probably won’t need UPS and its cooling requirement might extend to some passive heat-exchange or, in hot climes, an air-to-air heat pump.
We can expect developments in silicon, graphene, and others to kick off another unabated 20 years of Moores’ Law
The ‘edge’ facilities will start-off fairly small, maybe 3-5 cabinets (less than 10m²) in small rooms, either embedded in existing buildings or in what we would recognize as telecom shelters planted on city centre rooftops, but they will very quickly (in human terms) shrink to a 1m³ box in a cupboard or mounted on the pavement next to the traffic-lights – looking just like a broadband street cabinet of today.
They may increase in number and become denser, but each will continue to shrink in kW load and physical volume. They might start-off with deployments every 2km², but soon be found every 500m in densely populated city-centres.
Learning from the past
So, why do I think that? Well, as Churchill more-or-less said ‘learn history and never forget its lessons’. Our data centre history started in the mid-1950s and has been dominated by one feature – exponential growth in compute and storage capacity per unit of physical volume. Underpinned by Moore’s Law, the principle of which has found applicability in every ICT component, product and system in one form or another, we have seen ICT spread like a benign virus over the developed world.
Consider an ‘internet data centre’ at the turn of the century, housing 2,000 cabinets rated for 1kW load each. Just 18 years later that can be outperformed by a single cabinet housing a 5kW load, compaction of 2000:1 in floor space and 400:1 in power.
Over the same period, we have gone from precision cooling and tight humidity control to little more than opening the window – although the conservative data centre industry will take a few more years to get there. But the edge could easily be seen to use Class A4 (B or C) ICT hardware and need no thermal management at all in most climates.
Who is going to own the edge? I would bet on the telecom providers, both fixed-line and cellular mobile
There is little to no evidence that this technology capacity curve will slow down, in any case we can expect developments in silicon, graphene, and others to kick off another unabated 20 years of Moores Law.
Although the increased ICT demand does have a finite limit to the growth of the world population and their capacity to view UHD video on mobile phones, the IoT loads will be small data packets and growth will be dealt with by edge density – in other words more edge connection points, not fatter ones.
There is one further possible solution to power consumption from this ‘edge’ that is encouraged by the ‘micro’ access point I have described. If the edge is placed inside commercial buildings, particularly city centre hotels, then the single 5-10kW cabinet (maybe 1-2kW by 2030) could be liquid cooled, with the waste energy taken as hot water at 65°C that could be used by any hotel, 24/7. Thus, no power consumption by the internet.