Re: [E-impact] WindEurope Annual Event

[Jari Arkko <[e-mail]>]

Of course the variability of green power sources is an issue. To an extent we can deal with by varying consumption, e.g., turn the aluminum melters off when power is non-green or scarce, etc. Or data centers :-) 

Part of the problem is of course information sharing, to make sure consumption knows about the state of production. Here Internet helps, obviously :-) But of course there are also local means, e.g., a site that has its own green power knows how much it can consume, but even so, it may still need to communicate to parties it is serving, e.g., it may not accept new jobs unless there’s enough energy and so on.

Then a follow-up on this:

> In 5G we can have low and high power slices which are configured to consume high or low power respectively. For example, Low-power slices may use narrower channels, lower frequencies, and simpler antenna configurations to reduce energy consumption. Traffic is shifted to low or high power slices based on availability of renewable energy.
> 
> Does this work?

You can of course have network equipment configurations or virtual configurations that are geared towards a particular tradeoff. This is true of slicing as well.

However, for 5G perhaps a bigger question is what radios are on and how often and when they can sleep. The radio part is the largest fraction of energy use. But there’s a lot of radios, antennas, sites, coverage, even overlapping coverage and the usage can vary wildly. Think of city business district during work hours and a lot off demand, but more empty in the night when demand shifts to other areas. Potential energy savings don’t come from a particular user group but rather the aggregate. When we can design our system architectures so that the number of must-always-be-broadcasting-every-millisecond parts is minimized, that the overall number of radios on at all  can be scaled according to user demand, that equipment can fall asleep and wake up in extremely short time periods so that it can save energy when there are no packets to send, etc. And remember that packet transmission in mobile networks is highly scheduled/controlled which helps the equipment make well-informed decisions at least in short timescales.

Jari