INDEX: a status report

Size: px

Start display at page:

Download "INDEX: a status report"

Kelly Shannon Dean
6 years ago
Views:

1 : a status report May 1999 Pravin Varaiya, Richard Edell, Bjoern Rupp EECS, SIMS, Economics at UC Berkeley SIMS

2 What is? = Internet DEmand experiment Web page: Support: Cisco Systems, Pacific Bell, Hewlett Packard, and National Science Foundation. Page 2

3 What is our goal? Goal: Measure how much users are willing to pay for Internet Quality of Service (QoS). Dimensions of QoS bandwidth (symmetric and asymmetric) volume congestion delay to connect etc. Page 3

4 How do we do it? Give 150 Berkeley people "free" ISDN service hardware and setup costs are free we cover PacBell s monthly charges Offer them different QoS for different prices Simulate the QoS same QoS choices for several weeks different prices each week Note: real money, simulated service Page 4

5 Interface for : settings Page 5

6 Interface: Experiment Page 6

7 Interface: Pricing Page 7

8 Interface: choices Page 8

9 Interface: asymmetric BW Page 9

10 Interface: volume pricing Page 10

11 Interface: billing Page 11

12 Where we are Started providing service April 1, 1998 have about 70 subjects subjects proceed through experiments asynchronously Have run symmetric bandwidth asymmetric bandwidth volume pricing bandwidth + volume mixture (predetermined + self-selected) pay upfront for flat pricing Page 12

13 Symmetric bandwidth Usage overall: 8 kbs Note: 8 kbs is free 128 kbs 96 kbs 64 kbs 32 kbs 16 kbs Page 13

14 Scatter Plots price price 16 Kbps 32 Kbs u u32 Page 14

15 Scatter Plot, continued price 10 price Kbs Kbs u u96 price Kbs u128 Page 15

16 Demand estimates Reduced form: estimate demand as a function of price amount consumed = function of prices and individual characteristics Structural: estimate parameters of utility function choice of bandwidth depends on value of time, urgency, etc. as well as price and characteristics First look at reduced form, then examine some options for structural estimates Page 16

17 Log Regressions u128=-2.0p p p p p16 u96 =+1.7p p p p p16 u64 =+.77p p96-2.9p p p16 u32 =+.81p p p64-1.4p p16 u16 =+0.2p p p p32-1.3p16 Red = negative own price effect Light = not statistically significant Lesson: large negative own price effect, positive cross price effect Page 17

18 How good is the fit? R 2 ISE no ISE u u u u u ISE = "individual specific effect"; interpretation Conclusion: very good fits! Page 18

19 Structural estimates Notation b = bandwidth chosen x = bits transferred t = time at bandwidth b = x/b p(b) = price of bandwidth b Utility u(x) - [c+p(b)]t c is value of time; varies with circumstances f(c) is probability distribution of c Page 19

20 Choice Let (x,b*) be choice. Optimization requires: u(x) [c + p(b )] x b u(x) [c + p(b)]x b for all b. Rearrange to find min b <b p(b )b p(b)b b b c max b >b p(b )b p(b)b b b Note that this gives upper and lower bounds on cost of time. Monte Carlo studies look good. Page 20

21 Picture of estimation method total cost b 3 b 2 b 1 c L c U c Page 21

22 Illustration Histogram: nonparametric estimate Curve: normal approximation f(c) = frequency c = time cost Page 22

23 Volume experiments Two bandwidths 8 Kbs for free 128 Kbs for money Price volume downloaded prices range from 1 to 20 cents per megabyte monthly expenditures similar in bandwidth and volume experiments Page 23

24 Scatterplot Page 24

25 Log Regressions no ISE: u128 = p128 R 2 = 0.02 ISE: u128 = p128 R 2 = 1.00 Interesting finding: amount transmitted is not sensitive to price entire effect is individual-specific effect Why? Ticking clock effect? What does this mean for provisioning? Page 25

26 What s next? Current experiment: pay flat fee on Sunday to opt out of metered pricing Congestion pricing Simulate packet drop Charge for less congestion Move to ADSL or cable modem? What are you interested in? Page 26

Lecture 3: Modulation & Clock Recovery. CSE 123: Computer Networks Alex C. Snoeren

Lecture 3: Modulation & Clock Recovery CSE 123: Computer Networks Alex C. Snoeren Lecture 3 Overview Signaling constraints Shannon s Law Nyquist Limit Encoding schemes Clock recovery Manchester, NRZ, NRZI,