Part 7 of my series of comments on the New Scientist magazine series “Eight things you didn’t know about the internet” goes into the “carbon footprint” of the Internet: “Is the net hurting the environment?”, by Duncan Graham-Rowe.
It’s easy to think that the Internet can only be helpful to the environment. If you’re shopping online, you’re not driving to the store in your car. If you’re reading things online, you’re not printing the material, using paper and having trees cut down for it.
But all those servers, routers, and communication gateways take electricity to keep them running, and that means burning a lot of coal and splitting a lot of atoms.
Sending an email across the Atlantic Ocean does not burn any jet fuel, but the internet is not without its own, huge carbon footprint. One estimate suggests it takes a whopping 152 billion kilowatt-hours per year just to power the data centres that keep the net running. Add to that the energy used by all the computers and peripherals linked to it and the whole thing could be responsible for as much as 2 per cent of all human-made CO2 emissions, putting it on a par with the aviation industry.
At the technical plenary session in Vancouver in December 2007, the IETF looked at this issue (PDF), with guest speaker Bruce Nordman of the Lawrence Berkeley Lab — Mr Graham-Rowe quotes Dr Nordman’s LBL colleague, Jonathan Koomey.
Researchers in energy-efficient computers and networks are looking at ways to use less power in the first place, as well as at ways to power down components when they’re idle. Some approaches to the former are obvious: exchanging travel for electronic conferencing, developing devices that can do their jobs with less power, and so on. On the other hand, we’re countering those aspects with ever-faster processors and ever-larger hard drives, which take more power to run.
The challenge for powering down components is in accurately detecting — or anticipating — the need to use a component, and getting it powered-up and ready quickly. If we get it right, the benefit can be huge: a lightly used server may be spinning a hard drive 24 x 7, when it might be able to shut the drive down 50%, 60%, or even 80% of the time, if it can get it running quickly again. Perhaps a web server might have low-power flash memory that could hold the site’s home page, a few other key pages, recently accessed pages, and first-level links from those. Most of the time, it could serve requests from the flash memory and spin the disk up to prepare it for the next request, for which it’s more likely to need the disk.
With similar sorts of adjustments and increasingly sophisticated algorithms, we can reduce the power consumption of the different computer and network components, and really get the Internet to save natural resources.