Today I was asked to give a talk at the GreenTouch event in Milan. In this and in the following posts I will sum up the points I made. Hope they may stimulate some thoughts in this area that is one of those we are looking at at the EIT ICT Labs.
Consuming less or producing more?
Energy is a very complex area, and energy usage is, in a way, even more complex. In telecommunications, and more generally in ICT, Telecom Operators and IT Players are both bad guys and good guys. They are one of the most energy hungry sector and they fuel even more power consumption and at the same time they can provide tools for decreasing power consumption within their own networks and, even more importantly in the overall everyday life.
Technology evolution has made amazing progress in becoming more and more energy savvy, and yet energy use has exploded because of technology. Cultural aspects may be more important in energy savings than all sophistication in networks and data centres.
I want to take an "out of the box" view at this fascinating issue where we are all in agreement that the others should be more responsible and use less energy
The EIT ICT Labs is looking at Energy Aspects and has a dedicated area for this: Smart Energy. In addition on e of the High Impact Initiatives that has been defined focusses on the decrease of power consumption by networks looking at both innovative architectures for FRA (Future Radio Access) and to the world of IoT that through its numerical explosion will become one of the most important player in the energy area, both in terms of usage and in terms of innovative solutions.
A little story …
Power consumption has been an increasing burden to Telecom Operators. They started with the electromechanical systems where power consumption was basically a linear function of the traffic managed and then on to electronic systems where the power consumption is basically independent of the traffic. Whereas in the electromechanical systems the switch would require power to process the signal and select the routing in electronic system the power consumption is continuous. And the chips had to be cooled to the point that cooling them made up for 40% of the power consumption in a switch.
In the network the power consumption for transmitting bits has been decreasing thanks to optical fibre communications but at the same time at the edges the transmission of bits have become more onerous because it costs more, power-wise, to transmit through radio waves than through copper.
We’ve often heard the sentence that a bit is a bit, referring to the fact that we can associate to a bit a variety of meanings, like voice, image, clips, data and so on. However, a bit remains a bit even when its meaning is “nothing”, when nothing is associated to it. Hence, the power consumption is basically independent, at core network level (including switching and routing) from the semantics. Here it is not a matter of architecture, rather an issue of basic technology power consumption. And here, specifically for switches and places where there is a high density of electronics, the issue is heat dissipation that usually requires some sort of air conditioning that few years ago made up for 40% of the total power consumption.
On the Operator side, little can be done but to ask manufacturers for power savvy solutions and deployment leveraging on free cooling. Indeed, by including power and dissipation aspects in the call for tender Operators have stimulated Manufactures to come up with better offers from a power consumption and free cooling point of view.
Although the core network has increased its power consumption because of the continuity of the traffic, a bit is a bit…. and 0 is as power hungry as 1, the real voracity in power has shifted to the edges, both the broadband and the radio ones. If we look back thirty years ago, that is 1985, we had basically 0 consumption at the edges: wireless was non-existent and the copper drop was powered for very limited time, once it was used to talk (on the average 0.0125 Erlang, 18’ a day).
Then wireless came and then broadband came. And both are power hungry and … inefficient!
In radio you are spreading the electromagnetic field all around, even though you are making use of it in very tiny spots, in broadband you pump bits on a line that is wasting them most of the time.
Clearly, they are so inefficient that there is plenty of latitude for improvements.
Smart antennas with dynamic beam forming can dramatically improve the power budget as well as new ways of dealing with noise on an ADSL line. And these are just two examples.
The deployment of fibre can reduce the power budget, although architectures play a significant role here. Point to point fibres may provide unlimited bandwidth to the customer but they are much more expensive (also) in terms of power budget. On the other hand, GPON infrastructure can minimise the power budget on the switch side (but that is not so on the premises side…).
Indeed, the periphery is a crucial part of the network when we are interested in power consumption.
The network periphery has changed and quite a bit. When telecommunications started (1860ies) the phone was powered locally, by renewable energy that to a fractional extent even powered the switch. Then batteries where provided locally to save the cranking (we are lazy… same happened with the car). Then the operation cost moved batteries to a central location, by the switch and that was so for over a hundred years. And then cordless came, and answering machines, and faxes, and then cell phones, modems, smart Televisions…. and sensors. And … we have just began!
In the future we will be wearing wearable, pills will be active and communicating, ambient will be aware. All of this does not come for free, also from an energy budget standpoint. And keep in mind that periphery is both more pervasive than a network and less controllable than a network. As you move from network to periphery you are also shifting from engineering to culture.
Along with the changes of devices we have seen, and are seeing, a change in traffic patterns, as shown in the figure. If in the first part of telecommunications voice was dictating the traffic patterns, with an average of 3 minutes 6 times a day (also a consequence of tariffing – Operators used to readjust traffic to better fit the network capacity through tariffing scheme…), around the turn of the century most of the traffic -measured in bytes- was associated to big transactions (video). As we move into the next decade (beyond 2020) we are going to see a continuous growth of this kind of traffic and a faster growth of myriads of small, byte-wise, transitions as consequence of hundreds of billions of sensors.
As shown in the third part of the graphic voice represents a negligible part of the traffic measured in bits and in transactions. Note how the bit measure drives optimisation in transportation (technical solutions aiming at increasing signal transmission efficiency are good for power savings) whilst transactions wise power savings can be derived from better processing architectures.
This variety of traffics poses challenges to present network architectures and at the same times provides new options for increasing the power efficiency.
The IP protocol is not particularly efficient power-wise and this is a big issue when we deal with sensors spread in the ambient. New protocols are being studied specifically to address this issue and since we will be dealing with hundreds of billions of low power sensors (most of them scavenging energy) it is clear the importance on one hand of these studies and the overwhelming shift that we will have in the next decade towards non-IP communications. This will also involve different communications paradigms and FRA, Future Radio Access, is part of this new game.
One of the paradigm changes ahead will be a shift from connection driven by the network to connections driven by devices, something the Operators should better start thinking about.