Small Players Zhixue Lu 1 Prasun Sinha 1 and R Srikant 2 1 The Ohio State University 2 Univ of Illinois at UrbanaChampaign 1 Cellular Data Keeps Increasing 2 Mobile Data Increases more than 60 Annually ID: 313881
Download Presentation The PPT/PDF document "EasyBid: Enabling Cellular Offloading vi..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
EasyBid: Enabling Cellular Offloading via Small Players
Zhixue Lu1, Prasun Sinha1 and R. Srikant2
1The Ohio State University 2Univ. of Illinois at Urbana-Champaign
1Slide2
Cellular Data Keeps Increasing
2
Mobile Data Increases more than 60% AnnuallySmall Cells (Femtocells) Increase Spectrum ReuseSlide3
Femtocells: the Concept
Small in-home Cellular Base Station connects to the service provider’s network through owner’s broadband network
F
emtocell
Broadband Router
Internet
Core Network
Femtocell Gateway
3Slide4
Femtocells: the Facts
To Deploy Cellular Base StationsSite, Backbone and Power SupplyCostly to deploy7.9 Million Femtocells Deployed by 2013Almost all are residential and enterprise (small owners) FemtocellsAcquiring Access to these Femtocells is Important4Slide5
Proposed Incentive Mechanism: Auction
Why Auction? : Fair and EfficientTwo Types of AuctionsForward Auction: buyers bidReverse Auction: sellers bidConsider a Reverse Auction ModelBuyer: the wireless service provider (WSP)Sellers: the femtocell ownersReason: most owners have only one femtocell
5Slide6
Background
Desired Properties of AuctionsTruthfulness: bidders cannot get higher utility by lyingIndividual Rationality: utility of any bidder ≥0Common Auction MechanismsSecondary price auctionReserve price based secondary auction6Slide7
Imprecise Valuation: an Ignored Problem
Existing Works Assume Precise ValuationsValuations of Femtocell Owners Depend On:Cost of extra broadband traffic, electricity usageDegree of overload/delay toleranceWiliness to provide serviceMay vary over timeHard to Precisely Estimate
+
+
No Delay!
= ?
7Slide8
Assumptions
Sellers Can Estimate With Bounded Errors: True Valuation of f, Hidden Value
: Perceived Valuation of f, Exposed ValueDistribution of is knownTruthful Auctions: Sellers Submit Perceived Valuations Truthfully
8Slide9
Basic Form of Auctions in the Paper
Consider Reserve-Price based Secondary AuctionSecondary auction: truthful with precise valuationsReserve price: eliminate errors (uncertainties) in payments How It WorksConsider one seller a timeWSP sets a reserve price x
The Femtocell owner places its bidAuction succeeds and pay x to the owner if the bid ≤ xUtility of WSP is G-x, G: the savings of the WSP on each unit of data offloading9Slide10
Negative Utility of Femtocells
Femtocell Owners: Negative Utility when < Payment <
G=14,Uniform in [0,10] ,=2Reserve Price: x=$7: $8, : $6Negative utility: 7-8 = -1Individual Rationality Violated 10
0
8
10
6
4
2Slide11
Address Negative Utility Issue (Naïve)
The WSP sets a reserve price $6, payment $8Seller f wins and receives $8 if its bid ≤ 6Expected Utility of WSP: 3.6
= 3.6 Worst-case IR11
0
8
10
6
Reserve Price
4
Payment
2
=2
Slide12
Imprecision Loss
New Issue (Naïve): Imprecision LossFor Femtocell Owners:, No loss even if
,
Loss if
> 6
, Loss if
>
6
12
Imprecision Loss
(IL)
: Percentage of utility loss Due to Imprecision: 100%
No Imprecision Loss
0
8
10
6
4
Reserve Price
Payment
2
No Imprecision Loss
Slide13
Problem Definition
M sellers, distribution of valuations knownProblem: maximize Subject to: Sellers are comfortable to submit imprecise valuations
13Imprecision Loss
No Imprecision Loss
0
8
10
6
4
Reserve Price
Payment
2
No Imprecision Loss
1. The Worst-case Utility of
any seller ≥0
2.
Partial Truthfulness
:
percent do not lose any potential utility by submitting imprecise valuations
3.
Imprecision Loss
: The expected utility loss for each user (in red) is bounded (
)
Slide14
Solution: Multiple Reserve Prices
Example: 2-reserve-price Approach:
if
bid ∈ [0,4), approve and
pay $8
if bid ∈ [4,10], approve with probability 2
/3 and
pay $10 if it is approved
Truthful and IR with Precise Valuations
0
4
S
1
S
2
10
Payments:
P
i
Approval Ratios:
R
i
14
Segments:
S
iSlide15
Multiple Reserve Prices In Imprecise Valuation Auction
Two Reserve Prices
04
10
6
No Imprecision Loss
Imprecision Loss
No Imprecision Loss
S
1
S
2
15
WSP’s
Expected Utility
4.0 vs.
3.6 (Naïve)
Imprecision Loss
25%
vs. 100%
Percent of Sellers
in IL Range
40% vs. 40%Slide16
Algorithm Sketch
Input (Saving of WSP) (Estimation Error) Distribution of
(Constraints) Output$N$ Reserve Prices (Si, Ri, Pi, )Dynamic Programming based Algorithm: Pseudo-polynomial Time Complexity 16Slide17
Example
$8
Seller #1$6$1$3B
C
D
E
Seller #2
Seller
#3
Seller
#4
A
Seller
Seg
#
Ratio
Pmt
#1
S
1
1
8
#2
S
2
2/3
10
#3
S
2
2/3
10
#4
S
1
1
8
0
4
10
6
S
1
S
2
17Slide18
Simulation Result
Precise Valuation
Near Optimal
Imprecise Valuation
Increasing
Decreases
D
Decreases
18Slide19
Summary
EasyBid: A Reverse Auction Mechanism for Acquiring Access to FemtocellsIntroduce the notion of Perceived Valuation, Partial Truthfulness, and Imprecision Loss to characterize the quality of auctions with imprecise valuations.Present heuristic algorithms to maximize the WSP’s utility while satisfying given constraints on partial truthfulness and imprecision loss. 19