Keeping the Internet robust, innovative and democratic are the goals behind the Internet Measurement Lab (M-Lab), a project founded and supported by Google, the PlanetLab Consortium and the New America Foundation's Open Technology Institute.

Assistant Professor Nick Feamster and Associate Professor Constantine Dovrolis of the College of Computing's School of Computer Science serve on the steering committee for M-Lab. They are among a number of university researchers developing open-source Internet measurement tools for the M-Lab server platform, which is designed to advance network research and provide users with useful information about their broadband connections.

[Patients] wants security and privacy, but sometimes it’s hard for an individual to know what level of security and privacy they need.

M-LabOne such tool is NANO—Network Neutrality Access Observatory—created by Feamster and Professor Mostafa Ammar, with graduate students Mukarram bin Tariq and Murtaza Motiwala. NANO provides users with information "so they can determine whether their ISP [Internet service provider] is discriminating against a certain subset of users, applications or destinations," says Feamster.

In determining if an ISP discriminates—that is, if it engages in practices that degrade performance or connectivity for a service—NANO collects a broad range of performance data from clients around the world. NANO applies statistical techniques to infer whether an ISP is engaged in any form of discrimination while allowing for normal fluctuations in network performance, according to Feamster.

M-Lab tools would help support ‘Net neutrality

DiffProbe is a suite of tools that digs deeper into Internet performance to determine whether an ISP is classifying certain kinds of traffic as "low priority" and moving it through the slower portion of its network, resulting in delays, poor performance and other signs of inferior service. An ongoing project of Dovrolis and graduate student Partha Kanuparthy, DiffProbe non-intrusively probes the network path to diagnose the nature and extent of this kind of discrimination.


Computer Science professors Constantine Dovrolis (right) and Nick Feamster (above photo, at right) collaborated to produce M-Lab, a suite of tools that can measure the practices of Internet service providers.


Why would an ISP discriminate against certain kinds of traffic? Dovrolis explains with an example of an ISP that sells a Voice over Internet Protocol (VoIP) product. The ISP might move traffic from a similar free application (such as Skype) more slowly, while keeping traffic from its commercial VoIP application running through the faster part of its network—thus providing better service for the products it sells and degraded performance for its competitor.

The problem, for many people, is they believe ISPs should be "network neutral" and handle traffic irrespective of its content or application. Indeed, President Barack Obama pledged during his campaign to support legislation mandating ‘Net neutrality.

Network discrimination often invisible to users

Another form of discrimination is called “shaping,” which is the practice of automatically lowering a user's connection speed after a certain period of time. An ISP may trumpet a download speed of five megabits per second. But what it leaves in the fine print, if admitting to it at all, is that after the first megabyte, the bit rate automatically drops to one megabit per second.

ShaperProbe, one of the DiffProbe modules, looks for traffic shaping. "Shaping is hard to detect if you are a user," says Dovrolis, adding that the practice is "more common in Europe than in the U.S."

Yet another discriminatory practice addressed by DiffProbe deals with traffic quotas, where ISPs "allocate a certain amount of bandwidth for individual flows or individual applications," Dovrolis says. "Again, most users would not know this."

M-Lab tools may be downloaded as they become available at www.measurementlab.net. Data collected via M-Lab will be made available to the research community to allow researchers to build on a common pool of network measurement data.

"Our work doesn't take a stand one way or another about whether or not the ISPs should discriminate," Feamster notes. "What's important to that ISPs’ treatment of their users' traffic is transparent.

"If the ISP is implementing some kind of policy that discriminates against certain classes of users or certain classes of applications,” he says, “the users should at least know about it."