Ways to measure the energy cost of web pages…


Here’s some early results for questions I posted to UClue on estimating the energy needed to deliver a web page:

The first method:
Total energy use divided by web pages. This would work if we did the following:

  1. Calculate energy use by servers on the web host for a defined time. Divide by the number of bytes delivered to get a “cost per byte” delivery.
  2. Compute a relation between bytes of data and web pages – it would be different for each web page.
  3. Define “usage groups” of web pages – possibly there are distinct sizes of web pages, rather than a random distribution of sizes.
  4. Figure out the client composition. Impossible to get the specific model of the client computer, but we can differentiate Mac from Windows from Linux. We can also differentiate likely “old” computers by which OS version and browser they’re running – if it’s Netscape 4, it is likely ancient
  5. The problem is that many web browsers “lie” in their user-agent – one would have to research to find out which browsers could be trusted to say who they really are.
  6. Look at the location of traffic in the server logs. One could make broad estimates about the number of networks the message jumped through, and then estimate the electricity needed to run these networks. Divide by total network traffic. It may be that an “electro-packet” (need a better name) can be defined – the energy needed to route one packet from source to destination for different Internet network delivery scenarios.
  7. Add this all up to get a cost of delivery.

http://uclue.com/?xq=724

The Second Method – Bandwidth Divides Pages Delivered:
Estimate bandwidth consumed by a website over a time period, and divide by pages delivered. This doesn’t include the network and client rendering.

  1. Figure out the page views over a time period.
  2. Figure out the bandwidth usage by those web pages
  3. Figure out the cost of bandwidth
  4. Get the conversion factor – cost versus bandwidth – for that web host. Since most Web Hosts run close to break-even it will be a little more than the true cost of electricity, and has the advantage of implicitly including the entire cost for running the web host – back-end electricity use added
  5. Convert from cost to electricity via the conversion factor

The  Third Method – add it all up

  1. Figure how much power the whole Internet used over a time period. This would have to include the energy bill for all the hosting companies and network providers (the fraction for Internet traffic)
  2. Calculate the total electricity used by all client computers – reduce by the fraction of time they are not downloading/rendering a web page
  3. Divide by the number of web pages served

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s