Junseok Hwang, assistant professor in the School of Information Studies, and Stephen Chapin, associate professor in the L.C. Smith College of Engineering and Computer Science, received a $434,507 grant from the National Science Foundation to develop a service platform that would guarantee quality of service on the Internet by managing bandwidth using market principles.
The system would enable users and network service providers to reserve and trade bandwidth (network connectivity and capacity) based on their intended uses rather than having to compete for minimal bandwidth space over a crowded network. A key component of the system is the building of a middleware service platform to provide basic and critical functionality upon which service providers could build applications. To accomplish that, the researchers will develop specialized "bandwidth management points" (BMP) for the platform that would enable service providers to allocate a specific amount of bandwidth to users upon request.
"The Internet is congested and slow," Hwang says. "For general communication, it works. However, the Internet is unreliable for applications that include telephony and videoconferencing because of the inefficient way bandwidth is managed. This project is designed to address that problem."
Challenges involved in building the platform include security and inter-agent communication between BMPs across multiple domains of networks, Hwang says. Chapin will lead the research on the security issues, which include information privacy, integrity and the availability of service.
Hwang, Chapin, and a team of students from both the School of Information Studies and the College of Engineering will build a prototype of the system. They plan to test the feasibility of implementing it in the Internet 2 grid computing environments-Globus and Legion-to enable more reliable communication between computer clusters located at major research universities.
"The BMP we're working on is a building block that could be used to add better network scheduling functionality to the grid computing systems," Chapin says. "Currently, when users schedule computing time across a network, they are forced to rely on an error-prone process for allocation of communication resources. Users either simply assume that sufficient network capacity is available, or they rely on past history to estimate what resources might be available. The BMP system will enable users to schedule network capacity at the same time they reserve other resources."
