Evaluating Tradeoffs of Congestion Pricing for Voice Calls
Cached
Download Links
- [parapet.ee.princeton.edu]
- [sahara.cs.berkeley.edu]
- [sahara.cs.berkeley.edu]
- DBLP
Other Repositories/Bibliography
BibTeX
@MISC{Shih_evaluatingtradeoffs,
author = {Jimmy S. Shih and Randy H. Katz},
title = {Evaluating Tradeoffs of Congestion Pricing for Voice Calls},
year = {}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
Congestion pricing, varying prices according to the load, can more efficiently allocate scarce resources by modifying user behaviors. However, it presents operators with a tradeoff between system performance and user satisfaction. To evaluate this tradeoff, we propose a methodology that combines small-scale user experiments with simulations, and apply it to voice calls. We offered an IP-telephony service to 100 users over one year where each user is allocated a fixed number of tokens a week. Based on extensive experimentation, we found that the users are responsive to occasional price increases and that they can be enticed to accept congestion pricing. Using the results from the experiments, we developed a user behavioral model to simulate the effects of congestion pricing for larger user populations under the same token constraint. We proposed a set of thumb-rules operators can use to set the parameters for managing congestion pricing. By setting these to appropriate values, we estimated that congestion pricing can reduce call blocking rate by 60% or save provisioning by 20% while causing users to experience price changes in 4% of their usages. Our estimates strongly depend on our derived user model, thus we further verified the model by having real users using the service along with simulated users. Our findings indicate that congestion pricing can be very effective for voice calls because it only needs to be applied occasionally during high loads, and that users are tolerant and responsive to these occasional price increases.
Keyphrases
congestion pricing voice call occasional price increase user satisfaction high load token constraint user population user behavior small-scale user experiment extensive experimentation price change user model ip-telephony service real user simulated user system performance user behavioral model thumb-rules operator scarce resource fixed number
