That finding, along with new software for monitoring Internet video quality, could improve commercial videoconferencing services.
Better computer networks could also result, said Prasad Calyam, a doctoral student in electrical and computer engineering at Ohio State and lead author of the study. Calyam is also a systems developer at the Ohio Supercomputer Center (OSC).
The initial idea for the study was envisioned by Paul Schopis, director of network engineering and operations at OARnet. Calyam's other coauthors on the study include Mukundan Sridharan, a doctoral student in computer science and engineering at Ohio State, and Weiping Mandrawa, a network engineer at OSC.
Twenty-five sites in 14 different countries participated in remote videoconferences with Calyam's OSC lab for the study, and more than 500 traffic traces were used for the data analysis. Users rated the quality of their experience on a scale of 1-5 while the researchers inserted different amounts of delay, data loss, and jitter into the signal path. A score of 5 meant quality was excellent while 1 was very poor.
As the engineers expected, user ratings fell as the three negative factors increased. But they were surprised at how differently people reacted to each separate factor.
For long delays, the average score was 3.75, and for high rates of data loss, it was 2.25. But for large amounts of jitter, the users' dissatisfaction was more pronounced. The score quickly fell to 1, the lowest possible rating.
The fact that jitter emerged as the top pet peeve runs counter to conventional wisdom, Calyam said.
"We always thought delay was the primary factor that caused the video and audio impairments that people disliked the most, but that actually turned out to be the least important," he said.
He presented the study results at the recent Passive and Active Network Measurement 2004 conference in Antibes Juan-les-Pins, France.
Delay, loss, and jitter all depend on the delivery of packets of digital data over a network, Calyam explained. When packets are delayed, communication is slow; when they don't reach their destination, snippets of sound or video are lost.
Jitter is more complex, because it stems from the variability in packet delay. It is caused by network congestion, irregular packet sizes and packets arriving out of order. The signal becomes choppy, and the audio and video can run out of sync.
Between email, instant messaging, and Web browsing, the amount of traffic on the Internet varies day to day and by time of day, and that affects the quality of video transfers.
"The Internet has health diagnostics, measures of how well traffic is flowing on a given day," Calyam explained. "We wanted to see how some of the typical diagnostics affected perceptions of user quality."
Other studies have tried to gauge what makes people dissatisfied with videoconferencing, but this is the first study to obtain actual measures of how delay, loss, and jitter cause dissatisfaction.
Calyam and his colleagues were able to get this information by using a software program he designed called the H.323 Beacon. H.323 is a transmission protocol for real-time audio, video, and other data over networks such as the Internet.
Most people who use videoconferencing technology could benefit from using the H.323 Beacon, because the cost of a bad video feed can be high, Calyam said.
"Our research targeted high-quality videoconferencing systems used by CEOs and senators, not people using webcams in Yahoo chats," he said.
As popular as the Internet has become, experts have suggested that high-quality videoconferencing can only happen effectively over dedicated private networks or alternate networks such as Internet2.
Internet2 is run by a consortium of more than 200 universities working with industry and government to develop new networking technologies. Separate from the original Internet, Internet2 provides network bandwidth for research and education around the country.
One of Calyam's tasks at OARnet -- the networking division of OSC -- is to monitor the quality of data transfer over Internet2, and diagnose problems that cause delay, loss, and jitter. Schopis said that the H.323 Beacon does that job well.
"One of the more exhilarating aspects of the Beacon is it is applied research," Schopis said. "We can deploy it on the network and get meaningful results that can accelerate problem resolution."
Calyam said that comparable commercial tools cost as much as $250,000, but the code for the Beacon is open source, meaning anyone can download it for free and alter the code as they wish. It is also compatible with the Microsoft Windows operating system.
Even people who aren't network administrators can use the program, said Schopis.
"It enables effective communication for the end users, who often lack the experience or tools to communicate symptoms to technical staff," he said. "They can now capture test results and forward them to their network engineers."
The Beacon can be set to automatically contact system administrators when there is a problem, and it will also troubleshoot problems on the user's computer, such as a faulty sound card.
Calyam envisions that users could run the Beacon on the morning of a planned videoconference and get a kind of Internet weather forecast for that day. Should something go wrong during the conference, a system administrator could use the tool to diagnose the problem.
Calyam developed the Beacon to earn his master's degree at Ohio State. He is now working toward his doctorate, while he uses the program to help OARnet partners meet their requirements for conferencing and data transfer.
OARnet maintains the network that connects all of Ohio's colleges and universities. It also leads the Internet2 Technology Evaluation Center (ITEC), a consortium of universities, government agencies, and corporate supporters that examines critical new applications and infrastructure issues for Internet2.
The H.323 Beacon can be downloaded form the ITEC Web site: http://www.itecohio.org/beacon/index.html.
This study was supported in part by the Ohio Board of Regents and Internet2.
Additional contact: Paul Schopis
Schopis.1@osu.edu
614-292-1956
Written by Pam Frost Gorder
Gorder.1@osu.edu
614-292-9475