Regional Impact of Earthquake and Tsunami in Japan
Nov 7, 2011: Effects of the disaster on BitTorrent population correlates with earthquake and tsunami intensity.
As a follow-up to our previous article on the Tohoku earthquake and tsunami, we investigated the regional impact of the disaster on the BitTorrent population. We analyzed BitTorrent usage, along with reported earthquake and tsunami data for each prefecture to see if there was a correlation between the data sets.
We began by checking which regions saw the largest drops in users and as well as the regions that were most affected by the disaster. The graphs below show for each prefecture, the change in the BitTorrent population after the disaster (left), the intensity of the earthquake (center), and the height of the tsunami (right). To understand the impact on the BitTorrent population, we compare the number of users seen 24 hours before and after the disaster. In the BitTorrent population graph, a prefecture colored red means that the prefecture saw a drop in BitTorrent users and green means there was an increase. The brighter the color, the greater decrease or increase in BitTorrent users. In the earthquake and tsunami intensity graphs, the brighter the red means a stronger earthquake and a higher tsunami, respectively. The maps show that the BitTorrent population in prefectures closest to the epicenter (marked by the "X" on the maps) saw a greater decrease in BitTorrent users.
We quantified the relationship between BitTorrent and the disaster by performing correlation analysis. Here we found a a significant negative correlation between the drop in population and the height of the tsunami (r = -0.62), meaning the higher the tsunami, the larger the drop in the number of peers in that prefecture. We found an even stronger negative correlation between the logarithm earthquake intensity and BitTorrent impact (r = -0.81). This means that regions with higher earthquake intensity, tended to see greater drops in BitTorrent usage.
Japanese Earthquake and P2P Disruption
Mar 17, 2011: Despite the damage incurred by the earthquake last Friday, P2P usage in Japan was only temporarily affected.
Over the past week, multiple sources have reported that Japan's Internet services are still working, despite the damage suffered by some underwater communication cables during the Sendai earthquake and tsunami. To capture the extent of the impact from the perspective of our users, we looked at the number of online users before and after the event.
To see how BitTorrent usage was affected relative to typical activity patterns, we compared data from the week of the earthquake with data from the same time in the previous week. In the graph on the right, the yellow line represents the number of peers found between Wednesday March 2nd and Sunday March 6th, while the blue line represents the number of peers online between Wednesday March 9th and Sunday the 13th -- the week of the earthquake. For the most part, the number of BitTorrent peers that we see in Japan is similar between the two weeks.
The exception can be seen in the 24-hour period immediately following the earthquake, which occurred at 2:46PM local time on Friday, March 11th. As the graph shows, the number of BitTorrent peers found online differed by as much as 25% in this period compared to the previous week at the same time. Surprisingly, we still see a large number of peers online and activity returns to normal on Saturday morning.
We are carrying on additional analysis to better understand how these events affected not only Internet connectivity and performance for people in Japan, but also for users in other countries, as a result of the damaged communication cables and rerouting of network paths.
Impact of Virgin Media P2P Upload Throttling
Mar 11, 2011: P2P throttling trial cuts upload speeds by as much as 50% during peak hours of the day.
Virgin Media, an Internet service provider in the UK, recently completed a week-long test of its new upload bandwidth throttling policy that targets peer-to-peer traffic during peak usage times. Virgin plans to make the policy permanent if they deem the trial to be successful.
We find that this policy reduces users' upload bandwidth by as much as 50%, based on anonymized statistics contributed by P2P users inside the Virgin network.
On weekdays in the week before the throttling trial, users consistently used about 40% of their upload capacity. However, during the test (when P2P upload rates were throttled from 5PM to midnight), the maximum upload speeds we observed were significantly lower, sometimes only 20% of their capacity. The most significant reduction in speed occurs between 8PM and 10PM.
We also see that weekend upload rates were significantly reduced during the throttling period, from noon to midnight. Starting at about noon, our data shows a significant drop in upload speeds, from about 50% at 11AM to below 30% at 2PM. Later in the evening during throttling (at 10PM), upload rates fell to below 20% of capacity. In comparison, upload speeds before the trial averaged about 40% of capacity throughout the day.
(% of advertised)
Weekday upload speeds for Virgin Media customers in the week before ("No Throttling") and the week during the trial ("Throttling"). Shaded area shows the time when traffic management is enabled, from 5PM to midnight.
(% of advertised)
Weekend upload speeds for Virgin Media customers in the weekend before ("No Throttling") and the weekend during the trial ("Throttling"). Shaded area shows the time when traffic management is enabled, from noon to midnight.
