11 views

Uploaded by Jasmeet Singh

Game Theoretic Load Balancing

- Suspension System
- Games Managers Should Play
- Laurence J. O'Toole Jr -- The Theory–Practice Issue in Policy Implementation Research
- Behavioral Economics
- PBSProUserGuide10.4
- sg246649
- Spring Batch Docs
- TESIS BODY.pdf
- Re-Allocation
- A10_concurrency
- ie4902-set-2-5
- Problem Set 9
- grid
- D1,L9 Linear Programming.ppt
- Nash on the Crossroads of Information and Game Theory
- Lecture1-3-MEC223
- Influence of the Objective Function in the History Matching Process
- Parsons Structure of Social Action
- 1. Article (1)
- Task 8

You are on page 1of 35

in Distributed Systems

Anthony T. Chronopoulos

Department of Computer Science

University of Texas at San Antonio

San Antonio, TX, USA

atc@cs.utsa.edu

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

Load balancing: problem formulation

Load balancing

Given a large number of jobs, nd an allocation of jobs to

computers optimizing a given objective function (e.g. total

execution time or total cost).

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

Motivation for a game theoretic approach

Computational resources are distributed and used by many

users having dierent requirements.

Users are likely to behave in a selsh manner and their

behavior cannot be characterized using conventional

techniques.

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

Outline

Taxonomy of Load Balancing Approaches

A Noncooperative scheme

A Cooperative Scheme

A Dynamic Scheme

Application to Grid Models

Future Work

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

Categories of load balancing policies

Static policies: base their decision on collected statistical

information about the system.

Dynamic policies: base their decision on the current state of

the system.

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

Job Classication

Jobs in a distributed system can be divided into dierent

classes (multi-class or multi-user) based on their nature (e.g.

arrival rates, execution times etc).

So, the objective of the load balancing schemes can be to

provide

a system-optimal solution where all the jobs are regarded to

belong to one group (one class).

an individual-optimal solution where each job optimizes its

own response time.

a class-optimal (user-optimal) solution where the jobs are

classied into nite number of classes (users) based on their

nature and each user tries to optimize the response time of her

own jobs.

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

Load Balancing Approaches

I. Global approach

II. Non-cooperative approach

III. Cooperative approach

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

I. Global approach

Only one decision maker that minimizes the response time of

the entire system over all jobs.

Algorithms:

[Tantawi & Towsley 85][Tang & Chanson 00] nonlinear

optimization

[Kim & Kameda 92] ecient algorithm

[Li & Kameda 94] tree and star networks

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

II. Non-cooperative approach

Several decision makers (e.g. jobs, users) minimize their own

response time independently of the others and they all

eventually reach an equilibrium.

At the equilibrium a decision maker cannot receive any further

benet by changing its own decision.

This situation can be modeled as a non-cooperative game.

Solutions:

Wardrop equilibrium - innite # of decision makers.

Nash equilibrium - nite # of decision makers.

Algorithms:

[Kameda 97] Wardrop equilibrium;

[Roughgarden 01] Stackelberg game;

[Grosu 05] Nash equilibrium;

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

III. Cooperative approach

Several decision makers (e.g. jobs, computers) cooperate in

making the decisions.

Each of them will operate at its optimum.

Decision makers have complete freedom of preplay

communication to make joint agreements about their

operating points.

This situation can be modeled as a cooperative game.

Algorithms:

[Grosu IPDPS02] ,[Penmatsa IPDPS06] cooperative game

among computers.

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

Noncooperative Load Balancing Game among Users

([Grosu05])

We consider a distributed system that consists of n

heterogeneous computers shared by m users.

User j is a player and she must nd a load balancing strategy

s

j

= (s

j 1

, s

j 2

, . . . , s

jn

) that minimizes the expected response

time of her jobs.

The expected response time of user j is given by:

D

j

(s) =

n

i =1

s

ji

F

i

(s) =

n

i =1

s

ji

m

k=1

s

ki

k

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

The Distributed System Model

S1

S2

Sm

C1

C2

Cn

1

2

m

s11

s12

s1n

s21

s22

s2n

sm1

sm2

smn

1

1

1

2

2

2

m

m

m

Job

Assignment

Job

Assignment

Users

Computers Servers

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

Nash Equilibrium

Nash Equilibrium

A Nash equilibrium of the load balancing game dened above is a

strategy prole s such that for every user j :

s

j

arg min

s

j

D

j

(s

1

, . . . , s

j

, . . . , s

m

)

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

OP

j

: User j Optimization Problem

min

s

j

D

j

(s)

subject to the constraints:

s

ji

0, i = 1, . . . , n

n

i =1

s

ji

= 1

m

k=1

s

ki

k

<

i

, i = 1, . . . , n

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

Summary of results

The new algorithm (NASH) is compared with three other

existing load balancing schemes: Proportional Scheme (PS),

Global Optimal Scheme (GOS) and Individual Optimal

Scheme (IOS).

GOS minimizes the expected execution time for all the jobs of

all users in the entire system.

NASH nds the Nash equilibrium solution (i.e. it minimizes

the expected execution time for all the jobs of each user).

IOS nds the Wardrop equilibrium solution and PS is not

optimal.

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

The expected response time for entire system

(non-cooperative game approach)

0

0.05

0.1

0.15

0.2

0.25

0.3

10 20 30 40 50 60 70 80 90

E

x

p

e

c

t

e

d

R

e

s

p

o

n

s

e

T

i

m

e

(

i

n

S

e

c

)

System Utilization

NASH

GOS

IOS

PS

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

The expected response time for each user

(non-cooperative game approach)

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

Cooperative Load Balancing

We consider a distributed computer system that consists of n

heterogeneous computers (nodes) interconnected by a

communication network.

The load balancing problem is formulated as a cooperative

game among the computers and the communication

subsystem.

The Nash Bargaining Solution (NBS) is the solution for our

cooperative load balancing game which provides a Pareto

optimal and fair solution.

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

Cooperative Load Balancing Game

The cooperative load balancing game consists of:

n computers and the communication subsystem as players;

The set of strategies, X, is dened by the following

constraints:

i

<

i

, i = 1, . . . , n (1)

n

i =1

i

=

n

i =1

i

= , (2)

i

0, i = 1, . . . , n (3)

For each computer i , i = 1, . . . , n, the objective function

f

i

(X) = D

i

(

i

); for the communication subsystem, the

objective function f

n+1

(X) = G(); X = [

1

, . . . ,

n

, ]

T

.

The goal is to minimize simultaneously all f

i

(X),

i = 1, . . . , n + 1.

For each player i , i = 1, . . . , n + 1, the initial performance

u

0

i

= f

i

(X

0

), where X

0

is a zero vector of length n + 1.

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

Performance evaluation: Expected response time

(cooperative game approach)

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

10 20 30 40 50 60 70 80 90

E

x

p

e

c

t

e

d

R

e

s

p

o

n

s

e

T

i

m

e

(

s

e

c

)

System Utilization(%)

CCOOP

OPTIM

PROP

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

Performance evaluation: Fairness index

(cooperative game approach)

0.5

0.6

0.7

0.8

0.9

1

1.1

10 20 30 40 50 60 70 80 90

F

a

i

r

n

e

s

s

I

n

d

e

x

I

System Utilization(%)

CCOOP

OPTIM

PROP

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

Communication Time vs Expected Response Time

(cooperative game approach)

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.1

0.11

0.12

0.13

0.14

0.15

0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01

E

x

p

e

c

t

e

d

R

e

s

p

o

n

s

e

T

i

m

e

(

s

e

c

)

Mean Communication Time (sec)

COOP, 50% system utilization

COOP, 70% system utilization

CCOOP, 50% system utilization

CCOOP, 70% system utilization

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

Dynamic Load Balancing

Distributed dynamic scheme components:

Information policy: The number of jobs waiting in the queue

to be processed (queue length) is used as the state

information. This state information is exchanged between the

nodes every P time units.

Transfer policy: When a job arrives at a node, the transfer

policy component determines whether the job should be

processed locally or should be transferred to another node for

processing.

Location policy: If the job is eligible for transfer, the location

policy component determines the destination node for remote

processing.

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

Dynamic Non-cooperative Scheme with Communication

(DNCOOPC)

The goal of DNCOOPC is to balance the workload among the

nodes dynamically in order to obtain a user-optimal solution

i .e. to minimize the expected response time of the individual

users.

The derivation of DNCOOPC is based on the static NCOOPC.

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

Dynamic Global Optimal Scheme (DGOS)

The goal of DGOS is to balance the workload among the

nodes dynamically in order to obtain a system-wide

optimization i .e. to minimize the expected response time of

all the jobs over the entire system.

The derivation of DGOS is based on the static GOS similar to

DNCOOPC.

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

The expected response time for entire system

(dynamic load balancing approach)

0

0.05

0.1

0.15

0.2

0.25

0.3

10 20 30 40 50 60 70 80 90

E

x

p

e

c

t

e

d

R

e

s

p

o

n

s

e

T

i

m

e

(

s

e

c

)

System Utilization (%)

GOS

NCOOPC

DGOS

DNCOOPC

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

The expected response time for each user

(dynamic load balancing approach)

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

Job Allocation Schemes for Computational Grids

(GOSP and NASHP)

Grid is a type of parallel / distributed system which enables

the sharing, selection, and aggregation of geographically

distributed autonomous resources dynamically at runtime.

Computational grid: Tries to solve problems or applications by

allocating the idle computing resources over a network or the

internet

These computational resources have dierent owners who can

be enabled by an automated negotiation mechanism by the

grid controllers

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

Pricing Model

Players are the Grid Servers and the Computers

Reserved valuations

The server has to play an independent game with each

computer associated with it to form the price per unit

resource vector, p

j

.

In a system with m servers and n computers at time t, we

have m n bargaining games.

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

Non-cooperative Job Allocation Game

A Non-cooperative job allocation game consists of a set of players,

a set of strategies, and preferences over the set of strategy proles:

(i) Players: The m users.

(ii) Strategies: Each users set of feasible job allocation strategies.

(iii) Preferences: Each users preferences are represented by its

expected price (D

j

). Each user j prefers the strategy prole

if and only if D

j

(

) < D

j

(

).

Denition: A Nash equilibrium of the job allocation game dened

above is a strategy prole

(j = 1, . . . , m):

j

arg min

j

D

j

(

1

, . . . ,

j

, . . . ,

m

) (4)

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

Performance evaluation: Expected response time

(price-based job allocation)

0

0.05

0.1

0.15

0.2

0.25

0.3

10 20 30 40 50 60 70 80 90

E

x

p

e

c

t

e

d

R

e

s

p

o

n

s

e

T

i

m

e

(

s

e

c

)

System Utilization (%)

NASHPC

GOSPC

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

Performance evaluation: Fairness index

(price-based job allocation)

0.9

0.92

0.94

0.96

0.98

1

1.02

10 20 30 40 50 60 70 80 90

F

a

i

r

n

e

s

s

I

n

d

e

x

I

System Utilization (%)

NASHPC

GOSPC

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

Future Work

More results on load balanced job allocation in distributed systems

and validation by application to the Grid computing model.

Develop cooperative load balancing schemes for multi-class

jobs by taking the communication costs and bandwidth

constraints into consideration.

Develop dynamic cooperative load balancing schemes.

Extend the current non-cooperative load balancing scheme to

include the communication costs and bandwidth constraints.

Develop load balancing protocols based on mechanism design

that work in distributed systems shared by self interested

agents.

Implement the new schemes in conjunction with job allocation

schemes for grids.

Study the performance of the algorithms using existing

distributed systems simulation frameworks (e.g. SIMGRID).

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

References

D. Grosu and A. T. Chronopoulos, Noncooperative Load Balancing in Distributed Systems, Journal of

Parallel and Distributed Computing, 65(9), pp. 1022-1034, Sept. 2005.

S. Penmatsa and A. T. Chronopoulos, Cooperative Load Balancing for a Network of Heterogeneous

Computers, Proc. of the 20th IEEE Intl. Parallel and Distributed Processing Symposium, 15th HCW,

Rhodes Island, Greece, April 2006.

S. Penmatsa and A. T. Chronopoulos, Dynamic Multi-User Load Balancing in Distributed Systems, Proc.

of the 21st IEEE Intl. Parallel and Distributed Processing Symposium, Long Beach, California, March

26-30, 2007.

S. Penmatsa and A. T. Chronopoulos, Job Allocation Schemes in Computational Grids based on Cost

Optimization, Proc. of the 19th IEEE Intl. Parallel and Distributed Processing Symposium, Joint

Workshop on HPGC and HIPS, Denver, Colorado, 2005.

S. Penmatsa and A. T. Chronopoulos, Price-based User-optimal Job Allocation Scheme for Grid Systems,

Proc. of the 20th IEEE Intl. Parallel and Distributed Processing Symposium, 3rd HPGC, Rhodes Island,

Greece, April 2006.

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

Thank you

Questions ?

Anthony T. Chronopoulos University of Texas at San Antonio Game Theory Based Load Balanced Job Allocation

- Suspension SystemUploaded byradicri
- Games Managers Should PlayUploaded byapi-3831256
- Laurence J. O'Toole Jr -- The Theory–Practice Issue in Policy Implementation ResearchUploaded byintemperante
- Behavioral EconomicsUploaded bymrdanielb
- PBSProUserGuide10.4Uploaded byTalal AlFaisal
- sg246649Uploaded byvitorab
- Spring Batch DocsUploaded bysudhirsriramoju
- TESIS BODY.pdfUploaded byAngel
- Re-AllocationUploaded bynidee_nishok
- ie4902-set-2-5Uploaded byistenctarhan
- Problem Set 9Uploaded byforlionKing
- D1,L9 Linear Programming.pptUploaded bymokhtarppg
- A10_concurrencyUploaded byJoy Bagdi
- gridUploaded byajmal
- Nash on the Crossroads of Information and Game TheoryUploaded bysharmini
- Lecture1-3-MEC223Uploaded byjai
- Influence of the Objective Function in the History Matching ProcessUploaded byNeil Bryan Closa
- Parsons Structure of Social ActionUploaded bybarrosfelipetijuca
- 1. Article (1)Uploaded byAlin Chiper
- Task 8Uploaded bynavecg
- DifferenceUploaded bysaprsa1
- Managerial Decision Analysis On Goal ProgrammingUploaded byRich Kitten
- introduction-to-cloud-computing-1231468672808509-1.pptUploaded byŚoĦãīß MÀdnī
- 01 Intro(1)Uploaded byvinayrao92
- A Heuristic Task Deployment Approach for Load BalancingUploaded byJayaprabhu Prabhu
- A New Approach for Reactive Power Voltage Optimization Control of Regional GridUploaded byRVn Mor
- Movie MonitorUploaded byNaeem Gul
- CC Unit-IUploaded byshekhar_sk
- affisco2002Uploaded byBunty Evol
- 10.1.1.132.62.pdfUploaded byjorge_reeve

- why do we fall illUploaded byMonika Mehan
- fm1measphi0Uploaded byJasmeet Singh
- International_transition_Youth_wp.pdfUploaded byJasmeet Singh
- aaspUploaded byJasmeet Singh
- Using Nonlinear Kalman Filtering to Estimate Signals - Dan Simon(Excelente)Uploaded byBlack Fenix
- ekfUploaded byJasmeet Singh
- Z Transform(Q5 A1)Uploaded byJasmeet Singh
- ScbUploaded byJasmeet Singh
- Step by-Step 2012 WebUploaded byHuzaifa Shabbir
- IIT Quantization Cheat SlideUploaded byJasmeet Singh
- Facility Location Uni BonnUploaded byJasmeet Singh
- Anaheim Traffic OptimizationUploaded byJasmeet Singh
- MsaUploaded byJasmeet Singh
- k Centra Single Fac Loc Mat LabUploaded byJasmeet Singh
- Fairness IndexUploaded byJasmeet Singh
- MicEngExam TopicsUploaded byJasmeet Singh
- 02 Classification of SignalsUploaded byJasmeet Singh
- Delta FunctionUploaded byShaunak Patel
- Transmission Lines and Wave Guides Ec 1305Uploaded byainugiri
- Punjabi 2Uploaded byJasmeet Singh
- DotNet 2.0 License Agreement - EnglishUploaded byJasmeet Singh
- Kalman Dan SimonUploaded bydushtunini
- Introduction to Kalman Filter George WelchUploaded byJasmeet Singh
- Quantum ComputingUploaded bysoumo
- Deutsch D. - The Church-Turing Principle & Universal Quantum ComputerUploaded byrandomoptics
- Bayesian FusionUploaded byJasmeet Singh
- Covering LetterUploaded byJasmeet Singh
- hghghghUploaded byJasmeet Singh

- p5130 Lec09 Qualitative IvsUploaded byAhmedAlhosani
- 367345786 Netbeans Regula FalsiUploaded byArima Kousei
- 12_mathematics (Set - 1)Uploaded bySaksham Sharma
- OLGA_15_is_83710_psUploaded byAnonymous hdXKzBhoN
- Ccs c ManualUploaded byMichael Dixon
- Mechanical Design Data BookUploaded byvenugopalmanga
- Arduino Bots and GadgetsUploaded bypravin23105921
- BernheimCalculusFinalUploaded bysrinjoy
- edis 5320 lesson plan 1Uploaded byapi-340354632
- Model the Dynamics of Three-phase Round-rotor or Salient-pole Synchronous Machine - SimulinkUploaded byLanAnhThịNguyễn
- Lowest Common AncestorUploaded byRaqibul Islam
- GlossaryofSoftwareQATermsUploaded byapi-3834072
- Metodologías de MuestreoUploaded byGabrielSantelices
- On Syntropy & Precognitive Interdiction Based on Wheeler-Feynman’s Absorber TheoryUploaded byAnonymous 0U9j6BLllB
- Modeling Oxygen transfer of Multiple Plunging Jets Aerators using Artificial Neural NetworksUploaded byAdvanced Research Publications
- Micropile bearing plates.pdfUploaded byrowaters
- Handbook ResumeUploaded byPradisa Radja Lazuardi
- Ch-2- Strength of Materials by R.K.rajputUploaded byHusam Mustafa Mohamed
- A Compression on Parabolic and Skewed Mass DistributionUploaded byESSENCE - International Journal for Environmental Rehabilitation and Conservaion
- APL Basic Principles of Homing Guidance PalumboUploaded byÁlvaro Conti Filho
- Silt trapUploaded bymtrego9441
- ARRAY and Pointer.docxUploaded byRizwan Hameed
- CS 340 Databases Basit HafiqUploaded byZia Azam
- An Introduction to the Theory of StatisticsUploaded byRahul Biswas
- LATTICE BOLTZMANN SCHEME FOR HYPERBOLIC HEAT CONDUCTION EQUATION.pdfUploaded byMücahitYıldız
- Low Energy Electron DiffractionUploaded byBarathkumar Krishnan
- unit 2 examUploaded byapi-325599202
- Dialogical LogicUploaded byEricPezoa
- Research on Evolution Equations Compendium Volume 1Uploaded byJean Paul Maidana González
- MSA TrainingUploaded byshukumar_24