We’ve published network operator measurement results from World IPv6 Launch here. These results show some networks carrying impressive percentages of IPv6 traffic. The results are ranked by volume of traffic measured by Google. The percentages are a simple average of measurement results from multiple sources. Our measurement sources are Google, Yahoo! and Facebook and we have published results for all network operator participants for which we have at least two measurement sources and for which the simple average of those measurements is greater than 0.1%.
Folks interested in the measurement methodologies used by our sources should read on.
Google: We measure IPv6 adoption among Google users by adding measurement JavaScript to a random sample of visits to various Google properties. The methodology is similar to that described in our paper, Evaluating IPv6 adoption in the Internet. The JavaScript uses HTTP to fetch a URL from an IPv4-only hostname and a URL from a dual-stack hostname, in random order, and then reports the results. The measurement is attributed to the ASN that originates the user’s IPv4 address. Measurement endpoints use the same infrastructure that serve regular Google traffic.
Facebook: On June 6th, we examined page loads from randomly selected users. For each user we are able to test if they are using IPv6 to communicate with Facebook. We also know which ISP (World IPv6 Launch participant) they are using at their location. After collecting all data for June 6th, we calculated the percent of unique IPv6 users out of total unique users we saw from each ISP. Many millions of page loads were in our 24-hour data capture for these statistics.
Yahoo!: Yahoo!’s data is based on a sample of the user population hitting our “Front Page” sites (such as www.yahoo.com, ca.yahoo.com, etc). The percentage of IPv6 users is determined by an invisible image on the page being loaded from a dual stack (IPv4+IPv6) web server; the web server logs whether IPv4 or IPv6 was used. Each user selected for measuring IPv6 will only run this test to completion once per 24 hours, per browser session. We considered any ISP with more than 1000 users (extrapolated from sample set) to be measurable; we additionally considered any ISP with significant IPv4 users to be measurable.