User Experiments of Using Congestion Pricing to Allocate Access Link Bandwidth

Jimmy Shih
(Professor Randy H. Katz)
Packeteer

Congestion pricing, which varies prices according to load, more efficiently allocates scarce resources by modifying user behaviors. We study its effectiveness and user acceptance for allocating bandwidth at a local area network access link. We conducted two iterations of prototyping, deployment, and evaluation with about ten users. In the first iteration, we offered users three sizes of connectivity and varied their prices. We found that limiting user usages with different bandwidth sizes is not suitable for dealing with short bursts. In the second iteration, we offered users three levels of quality of service (QoS) that differed on degree of traffic smoothing and charged once every 15 minutes. We found that this scheme is effective because we can easily entice users to select a lower QoS, one with more smoothing applied to their traffic, by increasing the price for higher quality. Using a survey, users state that they would be willing to use the scheme because they only need to make a purchasing decision at most once every 15 minutes. Through simulations, we show that if half of the users in a large network can be enticed to have their traffic smoothed, then the burstiness at the network’s access link can be reduced by 20-30%.


More information (http://www.cs.berkeley.edu/~jshih) or

Send mail to the author : (jshih@eecs.berkeley.edu)


Edit this abstract