Wednesday, August 22, 2012

FCC Broadband Progress Report

The FCC released their eighth broadband progress report yesterday.
The most interesting part for me is when you get to page 45 and they start talking about actual adoption (in the US), as in the speeds that people are actually subscribing to, not what is available or offered.  Their buckets aren't all that granular and the data they used to build the report comes from June 2011 but they give you a good idea of what the spread looks like:
64.0% - At Least 768 kbps/200 kbps
40.4% - At Least 3 Mbps/768 kbps
27.6% - At Least 6 Mbps/1.5 Mbps

Effectively that means that 36% of the households where broadband is available do not subscribe to fixed-line broadband.  If we use the 64% that subscribe to at least some form of fixed-line broadband offering we get:
37% - Less than 3 Mbps/768 kbps
63% - At Least 3 Mbps/768 kbps
43% - At Least 6 Mbps/1.5 Mbps

With WebPagetest's default 1.5 Mbps/768 kbps DSL profile falling in the 37% of the population it is probably hitting somewhere around the 75th percentile.  Time to increase it to something closer to the median (say switch to the 5/1 Mbps Cable)?
I've generally been a fan of skewing lower because you will be making things faster for more of your users and you might be missing big problems if you don't test at the slower speeds but I'm open to being convinced otherwise.


  1. The shift to faster connections is certainly nice. But it's so easy for people to make the faulty leap that faster connections = faster page load. This is only half the story. Faster connections = larger bandwidth, but latency is the other side of the story. I wrote about this on the Zoompf blog earlier this year, in an appropriately named post "Bandwidth, latency, and the size of your pipe"

  2. +1 on latency. It isn't unusual with the 1.5Mbps DSL connection to see pages that are bandwidth constrained though and it all but goes away and everything becomes latency-constrained once you start going over 2-3Mbps (at least on the tests people are regularly asking for help looking at).

  3. I'm most interested in latency as it applies to mobile, where it is a tremendous problem. Sadly worse, CDNs rarely help mobile connections, since the telephone companies usually backhaul over their networks and dump it onto the internet somewhere that is far from your physical location. Unfortunately real world TTFB measurements for these mobile devices is difficult to measure. I don't know of any "mobile performance" test solutions that actual test with real mobile devices over a cellular network. Do you? Any studies about this? Can't find anything meaningful from Google on this.

  4. If you have a site with reasonable mobile traffic you can use the Navigation Timing support in Android 4 and later to get field data (or just grab the data from Google Analytics). I know we released some high-level aggregate stats from Analytics but I'm not sure if Mobile was called out (will take a look).

    The Akamai/Mobitest agents that can hook up to WebPagetest can run on carrier networks (and use real devices) but I don't expose any because of costs and highly variable results. Reach out to Guy and see if he has any private testing available.