2.5 million data points show: America's ISPs suck, and AT&T sucks worst


#1

[Read the post]


#2

And if you look at the pretty graph the worst performance is exactly when the heaviest load occurs. Could they spend more money and beef up their infrastructure, (and/or not oversubscribe their provisioning), of course, but that’s been true since the BBS days. How is this news?


#3

There’s been a lot of criticism of the study - the methodology wasn’t particularly open, or described in much detail in the article, and people who’ve tried to replicate it have had trouble - the choices of what to download from where don’t seem to be that good. (Source: lots of messages on Dave Farber’s Interesting-People mailing list, some of which disagreed with my summary.)


#4

And to make matters worse, out of a sampling of 99 countries, the US ranks 46 in cost/Mbps:

**The Blah:** Which leaves us with #46. Almost smack-dab in the middle is the US at an unimpressive USD $0.86/Mbps. Sure, it's slightly better than the UK at #48, but come on! This is the first country to put people on the Moon, yet they can't manage higher speeds at lower rates? You call that technology prowess? Pfft. Where are the freakin' rocket scientists when you need them? Russia (USD $0.24/Mbps) may not beat us in hockey, but they sure kick our butts in the long game.

The Cost to Connect - Internet Prices Around the World: valme.io/c/technology/vkqqs/the-cost-to-connect-internet-prices-around-the-world


#5

Hi Bill – I joined the IP list in response to the discussion, but since the exchange was largely the result of frustrating inaccuracies with the Guardian piece, I haven’t seen much contention after M-Lab’s blog post. One clarification: the blog post does not describe our methodology there, since it is exactly the same as our October 2014 Interconnection Study [1]. The intent was to refer readers to that document, which covers data access, server arrangements and analysis to a much greater degree than should end up in a short post.

I hope that covers all the outstanding concerns, however, there is the excellent question about duplication or challenges through other speed tests. The reality, is that others like Ookla are not wrong nor do they invalid the results against M-Lab – they both reflect broadband performance. The difference is generated by where the servers are located – in most cases Ookla’s sites seem to be close to the access ISP, often inside the network or in a CDN – it seems rarely in a transit ISP. Our sites are located almost always in major transit ISPs, and frequently we have multiple sites per city to cover as much infrastructure as possible – rather than attempting to infer the whole of the ISPs performance off one perspective. Moreover, single measurements make for bad comparative analysis – and we discourage attempting to determine congestion from one vantage point at one time. That’s where M-Lab’s dataset is powerful, it’s hundreds of thousands of consumers across the day and across different networks.

I would invite you to play with the data and test our findings, it’s all open and documented in the Interconnection Study.

[1] http://www.measurementlab.net/blog/2014_interconnection_report


#6

According to Level3, a major backbone, the US isps consider this a feature. They keep their internet connection overloaded, will not upgrade even for essentially free (L3 offered to pay for the hardware) so that they can force Netflix and the like to pay for private connections if they want customers to have a good experience.


#7

Yup. And that first chart, with Comcast and AT&T, are the only options in this huge city. (Some suburbs and other pockets get one or two other options, but they’re not a majority.)


#8

This test is totally bogus – It reports up/down speeds of 1.5 mbs; I get 270 kbs TOPS.


#9

You may be confusing bit and byte speeds. 1.5 Mb/s (megabits per second) is about 190 kB/s (kilobytes per second), and 270 kB/s is about 2.1 Mb/s. This test indicates the bit speed, while a browser might display the byte speed of downloads.


#10

5 1/2 hours to download a 2 hour movie rental (not HD) on iTunes?

No confusion.


#11

It looks good to me being on satellite, a bane of rural life. Max speed is about 4 Mbps for contiguous files, like FTP/SFTP. HTTP is not too bad at hundreds of Kbps. Mailto is awful and comparable or even worse than to dialup. I think it’s all the get requests up to the satellite that email requires. Still, all in all it is better than dialup which I had till a couple of years ago and a lot cheaper than using cellular for access. The main problem with satellite is bandwidth. My limit is 10 gigs before throttling and then drops to about 50 kbps—still better than dialup, but means I buy DVDs instead of using Netflix and similar services.

What happened to the rural internet access initiative?


#12

Congress rarely meets an unfunded mandate they don’t like.


#13

This topic was automatically closed after 5 days. New replies are no longer allowed.