Google uses integration to speed up the Web
By Murray Bourne, 27 May 2014
We all hate slow Web pages, and it's well-known that Google does too. In fact, when Google was new, I couldn't believe how quickly it would return search requests, measured in milliseconds.
Google's success as a company was built on optimization. They've put a lot of effort into making everything they do super fast.
Web page speed tests - area under curve
Google wants all pages to be fast, so they provide a free Web page speed test facility, WebPageTest.org. You can test any page there and see why it is so slow (or fast). It's great for those of us who are keen on speeding up our sites.
Of course, there are many things that will slow down a page, including slow Net connections, a slow server, size of the page (in kb), slow download speeds and slow browser processing speeds.
WebPageTest provides various test result measures, including connection time, download times and a metric called Speed Index.
Speed Index is an indication of how quickly the page elements appear on a page. In some cases, we see just a small part of the page and then see the browser "spinner" indicating more is to come (we are likely to leave in such cases). On other pages, we may see most of the page quickly, but the spinner indicates some more is coming.
Obviously the second cane is a better user experience, so it should get a better Speed Index score.
The Speed Index documentation explains how they calculate the score. It involves finding the area under a curve, which is a concept we learn when first meeting integral calculus.
They give a formula:
T is the "visually complete" time (total time to load and process the page); and
VC/100 is the "percent visually complete".
The areas are calculated at 0.1 s intervals (or similar) and added to produce the final index.
The area represented by the integral is the darker blue area in this chart, which represents a fast Web page (most of the content is visible quickly and the last 10% or so appears later).
[Image credit: WebPageTest]
So in fact, we are finding the area between 2 curves, the top one is the constant value 1 (or 100%), and the other is the VC curve.
Some closing ironies:
- The documentation page referred to above took around 27 seconds to "visually complete" status when I opened it today. They haven't even optimized images on that page to reduce the download time.
- Google is not always quick. Gmail has many elements and takes a long time to load when first opened.
See the 3 Comments below.