Where Does US Broadband Speed Really Rank?
This is the second part of pair of posts on US broadband speed; the first covered average download speed (33.5 Mbps in the US) and this puts that speed in context.
According to the Akamai State of the Internet report, the “Average Connection Speed” (which doesn’t mean what you think it means) in the US ranks 12th in the world in the third quarter of 2014, and according to the Ookla Netindex survey (AKA “Speedtest”), the average download speed in the US is 27th in the world, behind the Asian Tigers, Scandinavia, the Baltics, and several former Soviet satellites.
As I explained in the first part, there are differences in what Akamai and Speedtest measure: Akamai measures the load time for web page elements, mainly graphics, while Speedtest measures broadband capacity, commonly known as speed. When you run a Speedtest, you get an accurate assessment of your broadband pipe. But there are caveats.
First, the speed of your broadband connection doesn’t tell you much about your web experience because it’s only one of the potential limiting factors in our perception of web speed. The FCC explains this in its Measuring Broadband America 2014 Report (page 17):
In specific tests designed to mimic basic web browsing—accessing a series of web pages, but not streaming video or using video chat sites or applications—the total time needed to load a page decreased with higher speeds. However, the performance increase diminishes beyond about 10 Mbps, as latency and other factors begin to dominate. For these high speed tiers, consumers are unlikely to experience much if any improvement in basic web browsing from increased speed – i.e., moving from a 10 Mbps broadband offering to a 25 Mbps offering.
So your perceived web speed depends on broadband capacity, latency, and “other factors” such as server capacity. Latency is primarily a function of how far you sit from the web server. Because of some built-in quirks in TCP, we can download faster from web servers close by than we can from servers far away, all other factors being equal; this has to do with Slow Start, the mechanism TCP uses to prevent the Internet from collapsing from too much load.
Networks can’t get avoid congestion by adding capacity any more than you can outrun your shadow; the Internet is a packet-switching network, and such networks are designed to allow each user to consume the network’s entire pool of bandwidth for the duration of each packet. That’s the magic of the Internet.
Because we experience faster web page load times when the server is nearby, firms such as Akamai have a made a business out of making them close to everybody. When we read the New York Times online, the pages don’t come to us from a server in New York unless we happen to be in New York ourselves.
The gray lady is an Akamai content delivery network (CDN) customer, and that means that dozens – or perhaps hundreds – of servers around the US host the same content for the Times. If you happen to be close to one of these servers, you may very well see faster load time on a 10 Mbps connection than someone with a 100 Mbps connection will see 500 miles away. Akamai wants to know how fast we experience web pages loading, not simply how much capacity the pipe has. That’s their business, after all.
Internet users benefit from speed, but only to a point, as the FCC says. We also benefit from other parts of the Internet ecosystem, such as the density, performance, and price of CDN services. CDNs are a response to user demand for content, so we also benefit from heavy use by others. And we benefit from a business climate that makes it easy for companies to offer CDN services.
It’s interesting to look at the countries that rank below the US in the Akamai ranking and above the US in the Speedtest ranking. If my analysis is correct, this group of nations should have poorly developed CDN markets, low Internet use, and smaller volume of data consumption per user than the US does. Here’s the list of nations slower than the US per Akamai and faster per Speedtest by Speedtest rank:
Akamai excludes super-small nations, so Monaco, Aland Islands, Andorra, Jersey, et al., are excluded.
Now let’s look at data usage according to Bret Swanson’s analysis in “Internet traffic as a basic measure of broadband health.” Swanson takes Cisco data and averages it across actual Internet users to get averages for all traffic and for consumer traffic.
The Cisco data aren’t broken out by nation, but they do show that average data usage for US consumers is twice as high as it is for Japan and nearly three times as high as for Western Europe (2.7 times, to be precise.) The Cisco data also shows that Internet use in Eastern Europe and Asia as a whole is less than a third of the US level.
The conclusion is simple: The nations that have faster networks than the US according to Speedtest but slower ones according to Akamai have over-invested in broadband network speed and under-invested in Content Delivery Networks.
These networks are literally faster than their users need them to be, at least for the time being. While they may produce significant benefits in the future, it’s a certainty that US networks will also be faster in the future because they double in speed every two to three years.
So should we fret about our position in the global broadband speed rankings or should we celebrate the fact that the US dominates the Internet economy worldwide? It’s entirely possible that US networks are actually just as fast as they need to be for right now, if not faster. If this is the case, we should be wary of changing the policies that got us here.
I also provide a more in-depth analysis of these questions in my paper, “G7 broadband dynamics: How policy affects broadband quality in powerhouse nations”.
[Note: Since this post was written, Akamai has released an updated State of the Internet report covering the fourth quarter of 2014; there appear to be no significant changes. Washington Post’s GovBeat blog did some interesting journalism with Akamai’s data, showing where US states rank in relation to foreign countries.]