ISMA 2004 Workshop on Internet Signal Processing (WISP) 1 Perspectives on Resource Allocation Kameswari Chebrolu, Bhaskaran Raman, Ramesh R. Rao November.

Slides:



Advertisements
Similar presentations
“Students” t-test.
Advertisements

Hypothesis Testing. To define a statistical Test we 1.Choose a statistic (called the test statistic) 2.Divide the range of possible values for the test.
CPS Bayesian games and their use in auctions Vincent Conitzer
Ultimatum Game Two players bargain (anonymously) to divide a fixed amount between them. P1 (proposer) offers a division of the “pie” P2 (responder) decides.
Nash’s Theorem Theorem (Nash, 1951): Every finite game (finite number of players, finite number of pure strategies) has at least one mixed-strategy Nash.
3.1 Fair Division: To divide S into shares (one for each player) in such a way that each player gets a fair share. Fair Division: To divide S into shares.
Federal Communications Commission NSMA Spectrum Management Conference May 20, 2008 Market Based Forces and the Radio Spectrum By Mark Bykowsky, Kenneth.
Congestion Games with Player- Specific Payoff Functions Igal Milchtaich, Department of Mathematics, The Hebrew University of Jerusalem, 1993 Presentation.
Game Theory and Computer Networks: a useful combination? Christos Samaras, COMNET Group, DUTH.
MIT and James Orlin © Game Theory 2-person 0-sum (or constant sum) game theory 2-person game theory (e.g., prisoner’s dilemma)
Negotiation A Lesson in Multiagent System Based on Jose Vidal’s book Fundamentals of Multiagent Systems Henry Hexmoor SIUC.
Judgment in Managerial Decision Making 8e Chapter 4 Bounded Awareness
Excursions in Modern Mathematics Sixth Edition
Game Theory “Necessity never made a good bargain.” - Benjamin Franklin Mike Shor Lecture 13.
1 Unless otherwise noted, the content of this course material is licensed under a Creative Commons Attribution – Non-Commercial 3.0 License.
Game Theory The study of rational behavior among interdependent agents Agents have a common interest to make the pie as large as possible, but Agents have.
EC941 - Game Theory Prof. Francesco Squintani Lecture 8 1.
Competition Among Asymmetric Sellers with Fixed Supply Uriel Feige ( Weizmann Institute of Science ) Ron Lavi ( Technion and Yahoo! Labs ) Moshe Tennenholtz.
Todd and Steven Divide the Estate Problem Bargaining over 100 pounds of gold Round 1: Todd makes offer of Division. Steven accepts or rejects. Round.
Gabriel Tsang Supervisor: Jian Yang.  Initial Problem  Related Work  Approach  Outcome  Conclusion  Future Work 2.
Lecture 1 - Introduction 1.  Introduction to Game Theory  Basic Game Theory Examples  Strategic Games  More Game Theory Examples  Equilibrium  Mixed.
Dynamic Internet Congestion with Bursts Stefan Schmid Roger Wattenhofer Distributed Computing Group, ETH Zurich 13th International Conference On High Performance.
A Game Theoretic Approach to Provide Incentive and Service Differentiation in P2P Networks John C.S. Lui The Chinese University of Hong Kong Joint work.
6/2/2001 Cooperative Agent Systems: Artificial Agents Play the Ultimatum Game Steven O. Kimbrough Presented at FMEC 2001, Oslo Joint work with Fang Zhong.
Beyond selfish routing: Network Formation Games. Network Formation Games NFGs model the various ways in which selfish agents might create/use networks.
Information Gathering in Government Bailout Decision: An Experiment Ayung Tseng December 8,
Developing Principles in Bargaining. Motivation Consider a purely distributive bargaining situation where impasse is costly to both sides How should we.
Dynamic lot sizing and tool management in automated manufacturing systems M. Selim Aktürk, Siraceddin Önen presented by Zümbül Bulut.
Distributed Rational Decision Making Sections By Tibor Moldovan.
A Scalable Network Resource Allocation Mechanism With Bounded Efficiency Loss IEEE Journal on Selected Areas in Communications, 2006 Johari, R., Tsitsiklis,
Competitive Analysis of Incentive Compatible On-Line Auctions Ron Lavi and Noam Nisan SISL/IST, Cal-Tech Hebrew University.
Efficient agent-based selection of DiffServ SLAs over MPLS networks Thanasis G. Papaioannou a,b, Stelios Sartzetakis a, and George D. Stamoulis a,b presented.
1 Lecture 2: Negotiating Strategy Professor Keith Chen.
© 2007 Thomson South-Western. Consumers, Producers and the Efficiency of Markets Revisiting the Market Equilibrium –Do the equilibrium price and quantity.
L2: Market Efficiency 1 Efficient Capital Market (L2) Defining efficient capital market Defining the value of information Example Value of information.
1 A Cooperative Game Framework for QoS Guided Job Allocation Schemes in Grids Riky Subrata, Member, IEEE, Albert Y. Zomaya, Fellow, IEEE, and Bjorn Landfeldt,
The Agencies Method for Coalition Formation in Experimental Games John Nash (University of Princeton) Rosemarie Nagel (Universitat Pompeu Fabra, ICREA,
Game Theory.
Allerton 2011 September 28 Mathias Humbert, Mohammad Hossein Manshaei, and Jean-Pierre Hubaux EPFL - Laboratory for Communications and Applications (LCA1)
By: Gang Zhou Computer Science Department University of Virginia 1 A Game-Theoretic Framework for Congestion Control in General Topology Networks SYS793.
Efficiency Loss in a Network Resource Allocation Game Paper by: Ramesh Johari, John N. Tsitsiklis [ Informs] Presented by: Gayatree Ganu.
Capital Allocation Survey. Purpose of Allocating Capital  Not a goal in itself  Used to make further calculations, like adequacy of business unit profits,
NOBEL WP Szept Stockholm Game Theory in Inter-domain Routing LÓJA Krisztina - SZIGETI János - CINKLER Tibor BME TMIT Budapest,
A Hierarchical Model for Bandwidth Management and Admission Control in Integrated IEEE & Wireless Networks Dusit Niyato and Ekram Hossain IEEE.
1 Efficiency and Nash Equilibria in a Scrip System for P2P Networks Eric J. Friedman Joseph Y. Halpern Ian Kash.
Course Behavioral Economics Alessandro InnocentiAlessandro Innocenti Academic year Lecture 14 Fairness LECTURE 14 FAIRNESS Aim: To analyze the.
Multicast Scheduling in Cellular Data Networks Katherine Guo, Arun Netravali, Krishan Sabnani Bell-Labs Research Hyungsuk Won, Han Cai, Do Young Eun, Injong.
Lecture 2 Economic Actors and Organizations: Motivation and Behavior.
A Game Approach for Multi-Channel Allocation in Multi-Hop Wireless Networks Lin Gao, Xinbing Wang Dept. of Electronic Engineering Shanghai Jiao Tong University.
Fen Hou and Pin-Han Ho Department of Electrical and Computer Engineering, University of Waterloo, Waterloo, Ontario Wireless Communications and Mobile.
E VALUATION OF F AIRNESS IN ICN X. De Foy, JC. Zuniga, B. Balazinski InterDigital
MAP: Multi-Auctioneer Progressive Auction in Dynamic Spectrum Access Lin Gao, Youyun Xu, Xinbing Wang Shanghai Jiaotong University.
Dominant Resource Fairness: Fair Allocation of Multiple Resource Types Ali Ghodsi, Matei Zaharia, Benjamin Hindman, Andy Konwinski, Scott Shenker, Ion.
Introduction to Matching Theory E. Maskin Jerusalem Summer School in Economic Theory June 2014.
Todd and Steven Divide the Estate Problem Bargaining over 100 pounds of gold Round 1: Todd makes offer of Division. Steven accepts or rejects. Round.
Oligopoly. Oligopoly is a market in which a small number of firms compete. In oligopoly, the quantity sold by one firm depends on the firm’s own price.
Job scheduling algorithm based on Berger model in cloud environment Advances in Engineering Software (2011) Baomin Xu,Chunyan Zhao,Enzhao Hua,Bin Hu 2013/1/251.
Chapter 3 Fair Division.
Lecture 5A Mixed Strategies and Multiplicity Not every game has a pure strategy Nash equilibrium, and some games have more than one. This lecture shows.
NAREA Workshop Burlington, VT June 10, 2009 Yohei Mitani 1 Yohei Mitani Institute of Behavioral Science University of Colorado, Boulder Nicholas.
Incentives for Sharing in Peer-to-Peer Networks By Philippe Golle, Kevin Leyton-Brown, Ilya Mironov, Mark Lillibridge.
Beyond selfish routing: Network Games. Network Games NGs model the various ways in which selfish agents strategically interact in using a network They.
Jennifer Rexford Fall 2014 (TTh 3:00-4:20 in CS 105) COS 561: Advanced Computer Networks TCP.
3 SUPPLY AND DEMAND II: MARKETS AND WELFARE. Copyright © 2004 South-Western 7 Consumers, Producers, and the Efficiency of Markets.
Bargaining games Econ 414. General bargaining games A common application of repeated games is to examine situations of two or more parties bargaining.
SERENA: SchEduling RoutEr Nodes Activity in wireless ad hoc and sensor networks Pascale Minet and Saoucene Mahfoudh INRIA, Rocquencourt Le Chesnay.
Satisfaction Games in Graphical Multi-resource Allocation
Management support systems II
Value Based Reasoning and the Actions of Others
Excursions in Modern Mathematics Sixth Edition
Presentation transcript:

ISMA 2004 Workshop on Internet Signal Processing (WISP) 1 Perspectives on Resource Allocation Kameswari Chebrolu, Bhaskaran Raman, Ramesh R. Rao November 11, 2004

ISMA 2004 Workshop on Internet Signal Processing (WISP) 2 Resource Allocation Resource allocation is a critical issue in the design of communication and computing systems: –Bandwidth (e.g. wireless channel) –CPU (e.g. shared multi-user systems) –Memory, file cache Demand from multiple sources may exceed the supply –Uncertainty on the supply side (varying channel conditions) –Uncertainty on the demand side (fluctuating traffic) Mechanisms of control –Admission control, retransmission, differential pricing… Yet demand may exceed supply –How to divide the available resources? Proportional share is often taken to be the definition of fairness in this domain Is Proportional Share really fair? Are there other allocation mechanisms that deserve our attention?

ISMA 2004 Workshop on Internet Signal Processing (WISP) 3 Bankruptcy: –the liquidation value of a bankrupt firm is to be divided among its N creditors Taxation –the cost of a project is to be divided among N taxpayers Claims Problem Definition –An amount E has to be divided among a set of N agents with claims (c) adding up to more than E –A division rule R assigns an award (x) to each claimant such that 0 ≤ x ≤ c with awards adding up to E E c1c2 c3c4c5c6 x1x2 x3x4x5x6 Resource Allocation in other Domains

ISMA 2004 Workshop on Internet Signal Processing (WISP) 4 Proportional Rule (PROP) Rule : Explanation: Example : Make awards proportional to claims E = 150 c = [20,80,100] P = [15,60,75]

ISMA 2004 Workshop on Internet Signal Processing (WISP) 5 Constrained Equal Awards (CEA) Rule : Explanation : Example : Equality underlies many theories of economic justice Based on equality but respects upper bounds on awards Assign equal amounts to all claimants subject to no one receiving more than his claim E = 150 c = [20,80,100] P = [20,65,65] 150/3 = 50 [20,50,50]; 30/2 [20,65,65]

ISMA 2004 Workshop on Internet Signal Processing (WISP) 6 Constrained Equal Losses (CEL) Rule : Explanation : Example : Similar in spirit to CEA but focuses on losses claimants incur Assign equal amount of losses to all claimants subject to no one receiving a negative amount E = 150 c = [20,80,100] P = [3.33,63.33,83.33] 50/3 = 16 2/3 [3.33,63.33, 83.33]

ISMA 2004 Workshop on Internet Signal Processing (WISP) 7 Talmud Rule : Explanation : Example : Two regimes are defined based on the half-sum of the claims If half sum of claims is less than amount to divide, CEA is applied If more, every one receives half their claim and CEL is applied to the remainder Note that half-claims are used in the formula instead of claims E = 150 c = [20,80,100] P = [10,60,80] E = 150 c = [20,80,100] P = [10,60,80]

ISMA 2004 Workshop on Internet Signal Processing (WISP) 8 Talmud Continued Claim [20,80,100] and estate size is 150 First run CEA on half the claims against the full estate –Run [10,40,50] against 150 –This produces [10,40,50] with 50 leftover Then run CEL on remaining half claim against residual estate –Run [10,40,50] against a residual estate of size 50 –This produces loss of each in Round 1 or [0, 23.33,33.33] with a residual loss of 6.66 –Divide residual loss by residual claimants 6.66/2 = 3.33 –In round 2 claims reduce further to [0, 20, 30] Total award = [10,40,50] + [0, 20, 30] = [10, 60, 80]

ISMA 2004 Workshop on Internet Signal Processing (WISP) 9 Comparison of Rules a) Proportional d) Talmud b) Constrained Equal Awards c) Constrained Equal Losses

ISMA 2004 Workshop on Internet Signal Processing (WISP) 10 Properties of Rules Invariance under Claims Truncation: –If c i > E, replacing c i with E should not effect the chosen awards vector –Satisfied by CEA and Talmud, not by CEL, PROP Composition Down (Up): –When E is found to be less (more) than initially thought, the award vector should be the same for both of the following –Cancel initial division and apply the rule to the revised problem; or –Consider initial awards as claims on the revised E –Consider residual claim as claims on the increment in E –Satisfied by Proportional, CEA and CEL, not by Talmud No advantageous Transfer: –No group of agents should receive more by transferring claims among themselves –Only Proportional rule passes this test

ISMA 2004 Workshop on Internet Signal Processing (WISP) 11 Properties of Rules Continued CEA is the only rule such that for each problem –the gap between the smallest amount any claimant receives and the largest such amount is the smallest –the variance of the awards is the least for each problem CEA is the only rule that guarantees –Equal treatment of equals and –Invariance under claims truncation and –Composition up

ISMA 2004 Workshop on Internet Signal Processing (WISP) 12 Strategic Models Game 1: –Agents propose rules and the various rules are applied to the problem at hand –The claim of each agent is replaced by the maximal amount awarded to him by any one of the rules –The rules are applied to the revised problem and so on –This game has a unique Nash equilibrium which is the awards vector selected by Constrained Equal Awards (CEA) –no player benefits by changing their strategy if the other players keep theirs unchanged

ISMA 2004 Workshop on Internet Signal Processing (WISP) 13 Strategic Models Game 2: –Player A proposes an amount to other players in order –If any player B accepts his offer, B leaves –If a player B rejects the offer made to him, the next stage starts with B making an offer to the next player C; player A is moved to the end of the line –The process continues until only one player is left –No player should be offered an amount –greater than his claim or –the amount that remains to be distributed –As the discount factor of future utilities goes to one, the limit of payoff vectors in this game converges to Constrained Equal Awards (CEA)

ISMA 2004 Workshop on Internet Signal Processing (WISP) 14 Experimental Study on the Allocation Mechanisms Normative judgments: –what people decide to be “fair” in a dispassionate setting Actual negotiation: –when people negotiate among themselves to arrive at a mutually agreeable allocation Experiments have shown that people’s decision to allocate is: –Close to PROP in normative judgments –Close to CEA in actual negotiation Conjecture: –CEA is better suited to resource allocation –Leads to “better” user satisfaction

ISMA 2004 Workshop on Internet Signal Processing (WISP) 15 Comments and Conclusions Consider a communication link with rate 100 kbps –Suppose claims are 40 kbps and 160 kbps –PROP allocates 20 kbps and 80 kbps –CEA allocates 40 kbps and 60 kbps –In CEA, the 40 kbps demand is as important as the first 40 kbps demand of the larger 160 kbps claim How to Satisfy users? –Min-Claim-First satisfies maximum number of users, but starves larger claimants –PROP rewards inflated claims –CEA may make better sense Conclusions –It is debatable if proportional share is a good definition of fairness –A different set of rules advocated in bankruptcy claims problem –We advocate further study of CEA in networking given its interesting properties References –“Axiomatic and game-theoretic analysis of bankruptcy and taxation problems: a survey”, William Thomson. Mathematical Social Sciences, 45, (2003) –“Dividing Justly in Bargaining Problems with Claims”, working paper Gachter, Riedl 2004