The HTTP Archive crawls the world’s top 300K URLs twice each month and records detailed information like the number of HTTP requests, the most popular image formats, and the use of gzip compression. We also crawl the top 5K URLs on real iPhones as part of the HTTP Archive Mobile. In addition to aggregate stats, the HTTP Archive has the same set of data for individual websites plus images and video of the site loading.
I started the project in 2010 and merged it into the Internet Archive in 2011. The data is collected using WebPagetest. The code and data are open source. The hardware, mobile devices, storage, and bandwidth are funded by our generous sponsors: Google, Mozilla, New Relic, O’Reilly Media, Etsy, Radware, dynaTrace Software, Torbit, Instart Logic, and Catchpoint Systems.
For more information about the HTTP Archive see our About page.
I’ve made a lot of progress on the HTTP Archive in the last two months and want to share the news in this blog post.
- A major change was moving the code to Github. It used to be on Google Code but Github is more popular now. There have been several contributors over the years, but I hope the move to Github increases the number of patches contributed to the project.
- The HTTP Archive’s trending charts show the average value of various stats over time. These are great for spotting performance regressions and other changes in the Web. But often it’s important to look beyond the average and see more detail about the distribution. As of today all the relevant trending charts have a corresponding histogram. For an example, take a look at the trending chart for Total Transfer Size & Total Requests and its corresponding histograms. I’d appreciate feedback on these histograms. In some cases I wonder if a CDF would be more appropriate.
- We now plot the number of TCP connections that were used to load the website. (All desktop stats are gathered using IE9.) Currently the average is 37 connections per page.
- We now record the CDN, if any, for each individual resource. This is currently visible in the CDNs section for individual websites. The determination is based on a reverse lookup of the IP address and a host-to-CDN mapping in WebPagetest.
- Average DOM Depth
- Document Height
- Web pages are growing in many ways such as total size and use of fonts. I’ve noticed pages also getting wider and taller so HTTP Archive now tracks document height and width. You can see the code here. Document width is more constrained by the viewport of the test agent and the results aren’t that interesting, so I only show document height.
- localStorage & sessionStorage
- The use of localStorage and sessionStorage can help performance and offline apps, so the HTTP Archive tracks both of these. Right now the 95th percentile is under 200 characters for both, but watch these charts over the next year. I expect we’ll see some growth.
- Iframes are used frequently to contain third party content. This will be another good trend to watch.
- SCRIPT Tags
- The HTTP Archive has tracked the number of external scripts since its inception, but custom metrics allows us to track the total number of SCRIPT tags (inline and external) in the page.
- Specifying a doctype affects quirks mode and other behavior of the page. Based on the latest crawl, 14% of websites don’t specify a doctype, and “html” is the most popular doctype at 40%. Here are the top five doctypes.
html -//W3C//DTD XHTML 1.0 Transitional//EN http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd
html -//W3C//DTD XHTML 1.0 Strict//EN http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd
HTML -//W3C//DTD HTML 4.01 Transitional//EN http://www.w3.org/TR/html4/loose.dtd
Some of these new metrics are not yet available in the HTTP Archive Mobile but we’re working to add those soon. They’re available as histograms currently, but once we have a few months of data I’ll add trending charts, as well.
Big ticket items on the HTTP Archive TODO list include:
- easier private instance - I estimate there are 20 private instances of HTTP Archive out there today (see here, here, here, here, and here). I applaud these folks because the code and documentation don’t make it easy to setup a private instance. There are thousands of WebPagetest private instances in the world. I feel that anyone running WebPagetest on a regular basis would benefit from storing and viewing the results in HTTP Archive. I’d like to lower the bar to make this happen.
- 1,000,000 URLs – We’ve increased from 1K URLs at the beginning four years ago to 300K URLs today. I’d like to increase that to 1 million URLs on desktop. I also want to increase the coverage on mobile, but that’s going to probably require switching to emulators.
- UI overhaul – The UI needs an update, especially the charts.
In the meantime, I encourage you to take a look at the HTTP Archive. Search for your website to see its performance history. If it’s not there (because it’s not in the top 300K) then add your website to the crawl. And if you have your own questions you’d like answered then try using the HTTP Archive dumps that Ilya Grigorik has exported to Google BigQuery and the examples from bigqueri.es.
As part of the HTTP Archive project, I create MySQL dumps for each crawl (on the 1st and 15th of each month). You can access the list of dumps from the downloads page. Several people use these dumps, most notably Ilya Grigorik who imports the data into Google BigQuery. For the last year I’ve hesitated […]
Over the last week I’ve been investigating the cacheability of resources from Shopify. I would visit the page every few hours and track the number of 200 OK versus 304 Not Modified responses. To my surprise, Chrome’s Network tab indicated that almost all the responses were “from cache”. This didn’t make sense. In many cases the […]
Today I join Fastly as Chief Performance Officer. Read more in Fastly’s announcement. I’m excited to get back to the world of startups. People who have known me for less than 15 years don’t know that side of me. In my early career I worked at Advanced Decision Systems, Helix Systems (co-founder), General Magic, WhoWhere?, and […]