Study shows variations in Internet accessibility

A research team at the Viterbi School of Engineering’s Information Sciences Institute has discovered that the Internet “sleeps,” meaning that its accessibility varies over the course of the day.

The team consisted of USC alumnus Lin Quan, John Heidemann, a research professor and Yuri Pradkin, a USC researcher.

Countries in Asia, South America and Eastern Europe tend to have varied rates of Internet access over the course of the day compared to the United States, where people typically have consistent broadband access.

The team built on previous research by Heidemann and Pradkin that surveyed the relative size of the Internet. This outcome led them to study network outages and develop some new analytical approaches, leading to a piece of the data that showed the possibility of a “sleeping” Internet. This is especially important because it will serve in identifying the difference between a sleeping Internet and an Internet outage.

On Oct. 20, a video was released on Viterbi’s YouTube page further explaining the study through a more creative, animated approach. In addition, a peer-reviewed, technical paper will be presented in the first week of November describing the study. The data sets from the study have been made public in the hope that other people will build on the work.

“The longer-term direction of the work is using this information to understand how the different policy choices affect what’s going on and to understand how the diurnal is coordinated with other factors,” Heidemann said.

Data about diurnal networks — networks that vary in access and usage over the course of a day — have a multitude of uses. One use is in policy discussions, which make use of the quantitative aspect of the data when making decisions about free communication or economic growth. For example, the correlation between diurnal networks and GDP suggests an economic relationship.

“It’s not that diurnal [usage] causes low GDP, but if you want to increase your GDP, [a question to ask would be,] ‘Should you be investing in your network? If you’re investing in your network, should you be encouraging people to always have their Internet on or not?’” Heidemann said.

The team looked further into the relationship between GDP and the age of IP addresses. Though they did not find a direct relationship, they found that diurnal use is a reflection of the age of the IP address. Older IP addresses were used differently, because there were billions to give out. Now, however, IP addresses are used very carefully due to their limited availability.

In the immediate future, the data will most likely be used to crack down on network power outages, such as those in the New York area after Hurricane Sandy.

“We’re currently working with the FCC [Federal Communications Commission] to understand if our techniques are useful for understanding U.S. networks,” Heidemann said. “It turns out if you’re going to study network outages, you better know about sleeping networks because you don’t want to confuse a network that is underused relative to normal with a network that’s broken.”

Another use of the data, Heidemann noted, is to potentially investigate just how large the Internet is. By being able to count how many computers are on the network, the data would provide for further understanding of the baseline for the population, helping crack down on cybersecurtity. Diurnal information can determine if the estimate of the Internet size is accurately high or low depending on if it is day or night.

The team plans to present its research at a conference on Internet measurements on Nov. 5 in Vancouver, Canada.