Snowed down: Testing ISP Speeds on a Snow Day

What happens when the internet, the backbone of our high-tech economy, is down?  That’s a question recently brought to the forefront of the public’s attention, between the debate over Net Neutrality, hacking attacks, and President Obama’s recent proposal to expand broadband options.  Despite years of gains,  millions of Americans lack broadband access, and even where it is available speeds often lag other markets.  Yet we’re also more reliant on the web than ever before, constantly connected whether at home, at work, or anywhere on the go with smartphones. It’s easy to take for granted the benefits of fast, reliable broadband internet – at least until we lose our own connections.

Snow crossing signDuring the recent hurricanes and winter storms which managed to shut down major parts of the U.S. East Coast, there were numerous reports of internet service outages and website downtime for major hubs like Huffington Post and Netflix.  Some of these issues could have been caused by infrastructure damage from the storms, but at least some downtime was attributed to servers buckling under the load of millions more Americans simultaneously logging-on while stuck indoors during the inclement weather.  In fact, even during normal nights Netflix can account for as much as one-third of internet traffic during the peak evening hours, straining web servers and ISP networks.

Knowing these trends I formed a hypothesis: local ISPs would crumble during last week’s storm (Winter Storm Juno) under the heavy load of snowed-in users. To test my hypothesis, I recruited a simple network sample, emailing my friends across the New York metro area to run speed tests at the same time: around 9 PM EST on Monday, January 26th 2015. For the methodology we used to gauge speeds, while I collected details about each participant’s ISP and connection type, attempting to rule out as many variables as possible.  So here are some results from our little experiment:

Click the map's pins to see more detailed results from the "snow day" speed test by location

Contrary to my predictions, most of the volunteers* in my sample did not experience any noticeable slow-down or downtime.  By all accounts, internet speeds were very typical while New Yorkers were collectively snowed-in on that Monday night.   In fact the biggest variable had more to do with each network, depending on which internet service package each of my participants subscribed.

To be clear, I understand that my sample is not representative for every New Yorkers’ experience during this storm, but it was interesting to see how my personal experiences shaped my hypothesis; these results show that my own internet connection was among the slowest in the sample.  So perhaps I had been hitting my own ISP’s speed caps all along, rather than being slowed down by the heavy load on local networks.

Regardless of the results this winter storm was a unique opportunity to take a snapshot of the internet, even if it was just our little corner of the web.  Thanks again to everyone who helped collect data for this little experiment, most of whom asked to remain anonymous (you know who you are)!

Photo credit: “Snow Day STL” by Matthew Hurst, CC-BY-SA-2.0.

* Methodology note: most of the participants tested using their smartphone browsers (per my suggestion), which generally recorded slower speeds compared to using desktop/laptop browsers among those who tested both devices on the same network.  Another future test might check to see if ISPs increase their speeds for these “speed test” websites, similar to how some hardware manufacturers optimize their performance for benchmark testing.