Applying Congestion Pricing at Access Points for Data and Voice Traffic

Wednesday, October 30, 2002 - 17:30
TH 331
Jimmy Shih UC Berkeley
Congestion pricing, varying prices according to load, can more efficiently allocate scarce resources by influencing user behaviors. It can benefit users during congestion by giving them the option to receive good quality if willing to pay. In this talk, we investigate how to apply congestion pricing at well-defined bottlenecks like access points for both data and voice traffic. For data traffic, we conducted user trials of using dynamic pricing to allocate bandwidth at a local area network access link. The main challenge is that users are not used to deal with prices when using bandwidth. Using two iterations of prototyping, deployment, and evaluation, we found a scheme that is both effective and acceptable to real users. Through simulations, we showed that the scheme can significantly reduce the burstiness of a large network's access link. For voice traffic, we used a computer-telephony service to evaluate changing prices in the middle of phone calls. We offered the service to 100 users over one year to evaluate different pricing schemes. Changing prices is easier because users already pay by the minute when making calls. Through experimentations, we found that users are responsive and receptive to occasional price increases. However, each phone call only uses a fixed amount of resources, thus voice operators are really concern about its tradeoffs when involving thousands of users. Using simulations and user experiments, we showed that congestion pricing can greatly reduce call blocking rate or provisioning, while only requiring users to experience price changes occasionally.

Jimmy S. Shih received his B.S and M.E. degrees in Electrical Engineering and Computer Science, and B.A degree in Economics, from the Massachusetts Institute of Technology in 1997. He is currently a PhD candidate at the University of California at Berkeley. His research interests are in computer networks and distributed systems.