11 views

Uploaded by Jasmeet Singh

Game Theoretic Load Balancing

Game Theoretic Load Balancing

© All Rights Reserved

- Introduction to Advanced Research Methods Including Linear Programming, Dynamic Programming, and Game Theory_Brent M Eastwood
- IJCSIS Vol6 No1 October 2009
- Notes on Equilibria in Symmetric Games
- Milestones in the history of game theory
- Notes on Equilibria in Symmetric Games
- Distributed and Big Data Storage Management in Grid Computing
- Yale - Information Economics
- Bobby
- syllabus
- Minimization of Electrical Field in Stator Slot Groundwall Insulation of HV Generators
- 17 Manjula Shanbhog Research Letter Aug 2011
- MANOJ_KUMAR_MURLIDHARAN.C.V.
- vin-load balancing on grid
- slab design sheet
- Test Plans and Deliverables
- Eco Grid
- Paper 2012 b Parallel CFD
- set1-intro.ppt
- chandy
- A Game Theoretic Study of Attack and Defense in Cyber- Physic Al Systems

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

- Introduction to Advanced Research Methods Including Linear Programming, Dynamic Programming, and Game Theory_Brent M EastwoodUploaded byBrent M. Eastwood, PhD
- IJCSIS Vol6 No1 October 2009Uploaded byijcsis
- Notes on Equilibria in Symmetric GamesUploaded byshotorbari
- Milestones in the history of game theoryUploaded byvrimanek
- Notes on Equilibria in Symmetric GamesUploaded byhoseinkhazaei
- Distributed and Big Data Storage Management in Grid ComputingUploaded byijgca
- Yale - Information EconomicsUploaded byRoberto Sarkisian
- BobbyUploaded bydarkgirlslove
- syllabusUploaded byAnonymous bZTdTpL
- Minimization of Electrical Field in Stator Slot Groundwall Insulation of HV GeneratorsUploaded byIJEC_Editor
- 17 Manjula Shanbhog Research Letter Aug 2011Uploaded byclux
- MANOJ_KUMAR_MURLIDHARAN.C.V.Uploaded bymanojkumarmurlidhara
- vin-load balancing on gridUploaded byinterestingscribd
- slab design sheetUploaded byMuhammadWazimAkram
- Test Plans and DeliverablesUploaded byRomiSayago
- Eco GridUploaded byarsenic21
- Paper 2012 b Parallel CFDUploaded byJosh Strodtbeck
- set1-intro.pptUploaded bySiva Ramkrishna
- chandyUploaded bymishranamit2211
- A Game Theoretic Study of Attack and Defense in Cyber- Physic Al SystemsUploaded byMohammed Morsy
- 01- Introduction to Distributed SystemsUploaded byPeterArnold
- Course Handout - Dist. Systems 2019Uploaded byKanti Sudha
- Model for negotiation of refinancing gain from public-private partnershipUploaded byGibran Ius
- ClusterComp Published LibreUploaded bymohammedomar1974
- Clock SyncUploaded bykrishnab91
- Distributed Computing 2012-13(1)Uploaded byShubham Gupta
- 3G Radio OptiUploaded byVishal A Sheth
- InformationUploaded byTarikh Khan
- CS572Uploaded byFarooq Shaikh
- 10101002_devaUploaded byVarun Bhatt

- 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

- Fresher-Resume-Sample2.docUploaded byKarthiKeyan
- CommunicationsUploaded byjudesahayaraj
- G702_G802_G902_User_Manual_V1_0Uploaded byjcfdea
- 4. Css Assessment ProceduresUploaded byArchie Kins
- AIC+Uploaded byLovan So
- NWN Hordes Install GuideUploaded bytomat
- What is Table Maintenance GeneratorUploaded bySmith Kumar
- Cake Php CookbookUploaded bySarfraz Sheth
- T-REC-Q.850-201810-I!!PDF-EUploaded byBegumKarabatak
- 124171975 Modeling and Simulation of 3G UMTS in OPNETUploaded bykd kd
- IpsecUploaded byAlvarez Marco
- TEMS VUploaded byJuan
- Survey on Internet Access and Use by Filipino ChildrenUploaded bySherille Faith Marzon-Cartagena
- GA138 - SSP Technical Reference - 2.0.3Uploaded byNelson Parra Romo
- U90Ladder Special FunctionsUploaded byWorldtech Ascensores
- PacketFence Network Devices Configuration Guide-3.4.1Uploaded byluigyberto
- OPTICAL DISTRIBUTION FRAMES.pdfUploaded byDenisXavi
- FBL_WebCenter_Content_Integration.pdfUploaded byghazouanikarim
- SQL Connection to BWUploaded byrazakashif
- Nortel8010coUploaded byMichael Steven Warren
- How to Recover Vcsa 55 fUploaded byJesus Lopez
- oracle case studyUploaded byRagunath S Masthi
- SCC C6475 Catalogue[1]Uploaded byarash881
- Service Setup FailureUploaded byKeartichai Chomtaku
- Nokia Government Relations Policy Paper on 5th Generation of Communication NetworksUploaded byrainbows7
- NEPLAN-Tutorial-Elec-Eng-conejemplo.pdfUploaded byTIAGO3045
- Lab Guide UCCX 11.0 Enablement Lab v1Uploaded byCarlos Marcos
- Oracle OpenStack User GuideUploaded byAsit Swain
- Flt Lcp2 Specsheet FwbUploaded byAhmed
- obiee11gcmdUploaded byRavi Kumar Lanke