Jimmy Ssu-Ging Shih

EECS Department
University of California, Berkeley
Technical Report No. UCB/CSD-03-1235
2003

http://www2.eecs.berkeley.edu/Pubs/TechRpts/2003/CSD-03-1235.pdf

For alleviating network congestion, many researchers have advocated the use of congestion pricing, varying prices according to load, as a feedback mechanism for modifying user demand. However, it is not clear whether it can be designed to be acceptable to users and still be effective for operators. Thus we investigate user interface and system issues by applying congestion pricing at access points for voice and data traffic.

For voice, we deployed a computer-telephony service to 100 users for over one year to investigate changing prices during phone calls. We conducted user experiments to understand user response and acceptance to price changes. Using the results, we developed a user behavioral model to drive large-scale simulations for understanding how operators should manage congestion pricing and the tradeoffs they would face. The latter strongly depends on the user model. Therefore, we re-measured user reactions to price changes under a large-scale emulated service with many simulated users making calls and responding to the same price changes. We found that dynamic pricing can be effective for large user populations because users are receptive and responsive to occasional price increases. Prices only need to change 4% of the time, and doing so can dramatically reduce call blocking rate by 50% or save provisioning by 20%.

For data, we conducted user trials with 10 participants to examine using dynamic pricing to allocate bandwidth at a LAN access link. We found that offering users three classes of service based on traffic smoothing and charging once every 10-15 minutes is effective and acceptable. Through experimentations, users can easily be enticed to have their traffic smoothed by changing prices. Using surveys, users stated that they like choosing between different average performances and making purchasing decisions at most once every 15 minutes. Using simulations, we found that applying the scheme in a large network can easily reduce its access link bursts by 20-30%.

Through user studies and simulations, we show that applying congestion pricing at access points for voice and data traffic can be effective and acceptable. Thus, it should be considered for allocating limited bandwidth at access points.

Advisor: Randy H. Katz

BibTeX citation:

@phdthesis{Shih:CSD-03-1235,
    Author = {Shih, Jimmy Ssu-Ging},
    Title = {Applying Congestion Pricing at Access Points for Voice and Data Traffic},
    School = {EECS Department, University of California, Berkeley},
    Year = {2003},
    URL = {http://www2.eecs.berkeley.edu/Pubs/TechRpts/2003/5572.html},
    Number = {UCB/CSD-03-1235},
    Abstract = {For alleviating network congestion, many researchers have advocated the use of congestion pricing, varying prices according to load, as a feedback mechanism for modifying user demand. However, it is not clear whether it can be designed to be acceptable to users and still be effective for operators. Thus we investigate user interface and system issues by applying congestion pricing at access points for voice and data traffic. <p>For voice, we deployed a computer-telephony service to 100 users for over one year to investigate changing prices during phone calls. We conducted user experiments to understand user response and acceptance to price changes. Using the results, we developed a user behavioral model to drive large-scale simulations for understanding how operators should manage congestion pricing and the tradeoffs they would face. The latter strongly depends on the user model. Therefore, we re-measured user reactions to price changes under a large-scale emulated service with many simulated users making calls and responding to the same price changes. We found that dynamic pricing can be effective for large user populations because users are receptive and responsive to occasional price increases. Prices only need to change 4% of the time, and doing so can dramatically reduce call blocking rate by 50% or save provisioning by 20%. <p>For data, we conducted user trials with 10 participants to examine using dynamic pricing to allocate bandwidth at a LAN access link. We found that offering users three classes of service based on traffic smoothing and charging once every 10-15 minutes is effective and acceptable. Through experimentations, users can easily be enticed to have their traffic smoothed by changing prices. Using surveys, users stated that they like choosing between different average performances and making purchasing decisions at most once every 15 minutes. Using simulations, we found that applying the scheme in a large network can easily reduce its access link bursts by 20-30%. <p>Through user studies and simulations, we show that applying congestion pricing at access points for voice and data traffic can be effective and acceptable. Thus, it should be considered for allocating limited bandwidth at access points.}
}

EndNote citation:

%0 Thesis
%A Shih, Jimmy Ssu-Ging
%T Applying Congestion Pricing at Access Points for Voice and Data Traffic
%I EECS Department, University of California, Berkeley
%D 2003
%@ UCB/CSD-03-1235
%U http://www2.eecs.berkeley.edu/Pubs/TechRpts/2003/5572.html
%F Shih:CSD-03-1235