Google Inc. and two nonprofit partners Wednesday launched a Web site that lets consumers test their Internet connections to reveal possible interference and traffic management by service providers.
The site, Measurement Lab, addresses a need among academics who want to gather data on how Internet connections work in practice. While the workings of the core Internet "highways" are well known and standardized, it's difficult to find out what happens on the network of an Internet service provider, between the "highway" and the customer's home.
Internet service providers say they increasingly find it necessary to act as traffic cops on this stretch of the Internet to make sure that heavy users don't slow down their neighbors' connections. But the traffic management systems can have unintended consequences, and ISPs have been secretive about their workings for fear that subscribers will circumvent them. The Federal Communications Commission sanctioned Comcast Corp. last year for secretly stifling one particular form of traffic without telling subscribers.
One of the diagnostic tools on the M-Lab site, created by researchers in Germany, is specifically designed to detect interference of the kind that Comcast employed but has since abandoned.
Another tool is designed to detect whether certain types of traffic are being slowed. Cox Communications, the country's third-largest cable company, said this week it will test a system that temporarily slows some data to let more time-sensitive traffic through. That prompted skepticism from consumer groups that favor "Net Neutrality," which is the principle of equal treatment of Internet traffic.
Google spokesman Dan Martin said M-Lab's tools "could help users understand their connections, and would allow researchers to validate and explore what Cox is doing."
The search engine company will provide 36 servers in 12 locations around the world, and will cover the bandwidth costs. Its partners are the New America Foundation, a nonpartisan think tank, and the PlanetLab Consortium, which runs network experiments for researchers but lacks the capacity for large-scale testing by consumers.