Some web pages are designed to be power-hungry. They can afford to do that, since it is your power they are burning when you use your computer at home to view the web page.
If it is only a small amount of electricity, perhaps a watt, we might decide we don’t care. It might take days for 1 watt to turn into 1 cent of energy consumption. If it is more than this, maybe we do want to keep track. Apple famously redesigned its operating system and browser for advertising energy efficiency after discovering that more than half of the power consumption of a typical desktop computer can go to swap ads on web pages that the user isn’t even looking at. As Santeri Paavolainen noted in a recent post, “The question is how much power can a power-hungry website consume?”
Santeri went to the trouble of measuring the power consumption of isolated web sites displayed on a MacBook Pro. Notably, he did this with Flash turned off. That’s important because Flash is known to be power-hungry, too much so to run in web pages on smaller mobile devices. A single Flash advertisement can consume more electricity than an entire web page, so including Flash would allow the randomized selection of advertising to obscure the results. Santeri measured just 12 web sites, but found a wide range of power consumption. There was a consistent difference of nearly 10 watts between the more efficient web sites and the average. Just as important, one of the twelve sites consumed nearly 40 watts more than the average. We can assume this site suffered from sloppy coding, but it nevertheless demonstrates the benefit that we could obtain from a hypothetical power meter and on-off switch for web features.
Measuring the power consumption of a web page is a tricky thing to do. Even minor changes in browser, operating system, and hardware can change the power profile around. To cite a very obvious example, in the CRT era, web pages with a black background consumed less power than those with a brighter background, but this advantage disappeared when we switched to LCD and LED displays. Santeri had to control for various other factors and display only one page at a time to get meaningful measurements, and even then, the measurements are valid only for one browser version and one model of laptop. It seems safe to assume that power consumption would be higher across the board on a desktop system.
A Guardian article by Felix Salmon that focuses on the related metric of bandwidth points to commercial concerns as the cause of web page inefficiency:
. . . to a very large degree the owner of the website you’re visiting doesn’t actually control what you see, when you see it, how you see it, or even whether you see it. Instead, there are dozens of links in the advertising-technology chain, and every single one of them is optimising for financial value, rather than low-bandwidth user experience.
I know from my own investigation that this is true for the Shamanic Economist blog. The blog’s host goes to considerable trouble to keep track of page visitors, and that effort is the core of the web page you are looking at. Showing the post you are reading is an afterthought. The tiny advertisements at the bottom of the page and in the sidebar (on the desktop rendition of the page) routinely take up more bandwidth and presumably consume more power than the content of the page.
None of this will change as long as we don’t have meters for web page bandwidth and power. If a web site freezes and crashes frequently, you can guess that it is sucking bandwidth and power in a big way, but so do more carefully tested sites that look perfectly stable. If you are a web coder, you know how to examine the components of a web page you visit, but even most web developers aren’t coders, so they don’t necessarily have a way to measure the results of their work.
The bottom line: There is plenty of room for improving efficiency in the delivery of web content, but it is hard to see where those changes will come from as long as monetization remains the driving focus of the web.
