Georgia Tech Study to Gauge Internet Performance for Computer End-Users

Georgia Tech Seeks Research Volunteers for NETI@home: The Search for a Faster Internet

Ever notice your Internet connection seems slow? Ever wonder what's causing the slow down? You are not alone. Millions of users bemoan sluggish downloads and slow email but rarely know the cause of the delays.

Researchers at Georgia Tech have developed a technology to find out the how the Internet is performing from the "regular" end-users' perspectives. With this information, they can design and develop network solutions to relieve these bottlenecks. To do this, they need volunteers for the NETI@home project, which stands for "network intelligence."

Currently industry and academia use data on the Internet's performance measured at various router points out in the Internet, before it reaches the individual user (see http://weather.uci.edu). Georgia Tech researchers think a better approach is to find out how the Internet is performing from the user's point of view-at each personal computer.

"We think a better solution is to measure performance at the individual user level to determine what affects Internet traffic, but currently this data doesn't exist. That's where NETI@home comes in," says George Riley, professor of electrical and computer engineering and adjunct professor in the College of Computing. "We need thousands of computer users to use our free NETI@home software to help us gather this data."

To do this, Riley and graduate student Robby Simpson developed an open source software application that collects network performance statistics such as average response time, average round trip time, connection times, download times, and number of packets and bytes sent and received. The application then regularly reports these statistics to the NETI@home server at Georgia Tech. NETI@home is designed to be an unobtrusive software system that runs quietly in the background with little or no intervention by the user. The reports sent to Georgia Tech are also stored on the user's computer, so the user can see what statistics are gathered.

NETI@home users select a privacy level that determines what types of data will be gathered, and what will specifically not be reported. The application currently offers three privacy levels- low, medium, or high-allowing the user to choose the level of detail of the information collected, such as fully reports IP addresses, reports only the network portion of the IP address or does not record any IP addresses.

The researchers will collect the data and make it available to other network researchers. They expect the data will show trends of Internet performance or indicate security problems such as a huge spike of activity that might be caused by worms.

"Among the other things we anticipate this data will help us with is to give a reasonable picture of what Internet users demand of the network during normal activities. This will lead to better simulation models of Internet users behavior, leading to more accurate simulations. For example, with our NETI@home data, we might observe that a typical user experiences an average round trip time of 100 milliseconds and traverses on average 8 Internet routers. We would then construct our simulations to choose connection endpoints that have similar metrics," says Riley.

Computer users around the world are encouraged to download the free application to their computer, to help researchers in the quest for improved Internet performance. At www.neti.gatech.edu, users will find detailed instructions and answers to questions about the NETI@home effort. NETI@home participants also may receive an optional NETI map, a mapping application that displays a geographical map of where their computer is connected on the Internet.

In April the research team will present a paper about NETI@home at the international conference PAM 2004 Passive & Active Measurement Workshop, to be held near Nice, France. To download or learn more about NETI@Home, visit www.neti.gatech.edu.