The IETF duty for Carbon emissions, energy use &ct.

I think that whole conversation is missing the point. While one conference with one and a half thousand participants has a significant carbon footprint the more significant issue is the net impact of Internet technology. We can do far more by enabling changes in the behavior of the billion plus Internet users than our own use.

Whether your objective is to reduce carbon emissions because of global warming concerns or reduce dependence on fossil fuels that are increasingly in short supply, there is a real incentive to reduce energy use. 

Here are some options:

1) Teleconferencing that works.

If teleconferencing actually worked the need to hold face to face meetings would be reduced. By work I mean really work, not almost, not provided there is no NAT, not provided the firewall has pinhole router configuration. Today this is simply not the case, in fact it is pretty much impossible to get the software providers to even tell you which ports they use. For fun try to find the information on Microsoft's site.

This one is in scope for the IETF.

2) Enable better power management. 

While I do not believe the claims that the Internet takes 9% of the US energy supply, data centers are increasingly power hungry and the cost of electricity and equally important, cooling is a major factor in decisions to locate new data centers. Despite this, most data centers operate far below peak capacity and even when a data center is operating at 'capacity' the majority of the actual servers are not, only the ones that are the bottleneck.

Make it possible to spin up additional machines as needed to respond to increases in demand and standby machines can be powered down.

Certain approaches are in scope for IETF, others are not.

3) Removal of waste heat

Today most data centers use active cooling, albeit in a pretty inefficient fashion. Air is cooled to room temperature using refrigeration equipment, blown through the racks and the waste heat expelled into the atmosphere. So we use vast amounts of energy to make heat and then use more energy to vent it to the atmosphere.

First problem here is that air is not an efficient means of heat transfer, its loud and you have to use air at human-comfortable temperatures. This means that your racks are always going to be warm. Cool racks are more reliable than hot ones.

Passive technology such as heat pipes are much more attractive. I believe that future data centers will be built around equipment racks that contain separate busses for low voltage power, network and heat elimination. The heat busses will connect together to form a cooling grid for the machine room which will in turn vent to the outside air. Rather than using energy to expel the heat the process could be used to extract energy, even though the heat is low grade (below boiling point of water at atmospheric pressure) it is clean.

Getting the temperature down below the external ambient temperature would require an energy input of course, that might make sense if you were going to go to a really low temperature and overclock your way out of a performance crunch (i.e. liquid nitrogen temperatures).

This one is out of scope for the IETF, in fact its pretty much out of scope everywhere at the moment, its one of those issues that falls between the cracks of mechanical engineering and electrical. There are plenty of folk who think about these issues on the small scale but rather fewer thinking about what a data center built around these ideas would look like.