Sujit Dey

Adaptive Applications for Wireless Information Technology

Adaptive Applications for Wireless Information Technology

Sujit DeySujit Dey

ECE Department

University of California, San Diego

dey@ece.ucsd.edu

http://esdat.ucsd.edu

Sujit Dey

Ubiquitous Mobile Information TechnologyUbiquitous Mobile Information Technology

In-home Network

MSC

BS

The Internet

OpticalNetworking

Ad-hoc SensorNetworks

On-bodynetworking

Heterogeneous Network Configurations, Access Technologies, Data Services, and Network Appliances

Sujit Dey

Applications & Data Services for the Communication Era

Applications & Data Services for the Communication Era

• Computing Era– Primary Resource: Computing

– Paradox: Computing resources increased over years, but so did problem complexity

– Solution: Efficient computing algorithms: minimize computation ops

– Metrics: Runtime, space complexity

• Communication Era– Primary Resource: Bandwidth– Paradox: As bandwidth capabilities increase, so will new data-intense

wireless services and number of users– Solution: Adaptive Applications: minimize data traffic– Metrics: Quality of data; Quality of service

Sujit Dey

Our Approach: Adaptive Wireless DataOur Approach: Adaptive Wireless Data

The Internet

Base Station

Wireless GatewayServer

Web Servers

ConfigurableHandheld Client

BrowserApplication

ShaperApplication

Shaper Network ConditionsNetwork

Conditions

ApplicationRequirementsApplication

Requirements

Access Profile

Access Profile

Client Characteristics

Client Characteristics

Client Characteristics

Client Characteristics

Sujit Dey

Energy-Constrained Wireless Application Design : Motivation

Energy-Constrained Wireless Application Design : Motivation

Start

BreakingNews

Politics Sports Enter. Finance

BNHeadlines

PoliticsHeadlines

FootballHeadline

s

BasketballHeadlines

Science

Fin.Headlines

Enter.Headlines

ScienceHeadlines

BNSummary

PoliticsFull Story

PoliticsSummary

ScienceSummary

Weather

ScienceFull Story

FootballSummary

FootballFull Story

BasketballSummary

BasketballFull Story

Enter.Full Story

Enter.Summary

Fin.Summary

Fin.Full Story

0.30.1 0.1 0.2 0.05

0.150.1

10.5

0.5 1 0.40.20.3

0.9

0.1

0.6

0.4

0.8

0.3

1.0

1.0

0.2 0.9

0.1 0.4

END

0.4

0.9

0.3

0.1

0.5

Sujit Dey

Wireless Application Design : Energy OptimizationWireless Application Design : Energy Optimization

Politics Sports Enter. Finance

PoliticsHeadlines

FootballHeadline

s

BasketballHeadlines

Science

Enter.Headlines

PoliticsFull Story

ScienceFull Story

FootballSummary

FootballFull Story

BasketballSummary

BasketballFull Story

Enter.Full Story

Enter.Summary

Fin.Full Story

0.30.1 0.1 0.2 0.05

0.150.1

0.50.5 1 0.40.2

0.3

0.1

0.4

0.3

1.0

1.0

0.2

0.1 0.4

END

0.4

0.9

0.3

0.1

BreakingNewswithSummary

Start &Weather

Fin.Headlines&Summary

0.5

PoliticsHeadlines

PoliticsFull Story

Headlines

Sujit Dey

Wireless Applications : Energy Modeling & Optimization

Wireless Applications : Energy Modeling & Optimization

Goal : Develop energy modeling methodology for wireless application design exploration and optimize energy consumption

Steps : • Develop an energy model for handheld clients

(PalmVII) using measurement based methodology• Develop a generic energy estimation methodology

for wireless applications

Sujit Dey

Energy ModelEnergy Model

EncryptionEncryption

Connection Setup

Connection Setup

TransmissionTransmission

DecompressionDecompression

DecryptionDecryption

ReceptionReceptionIdlingIdling

TimeTime

TimeTime

T

T * Enc

L

R

R / Comp

R * Enc / Comp

E(R,T,C,L,Enc,Comp) = C0

+ Cidle*L

+ Cr*(C*R*Enc/Comp)+

+ Ct*(T*Enc)

+ Cenc*T + Ccomp*R

Sujit Dey

Validation of Energy ModelValidation of Energy Model

• Comparison of measured energy consumption with energy estimation using the proposed model– Based on average energy consumption over 10 accesses

Energy Model : Sufficiently Accurate to drive application level energy optimization

Application Measured Energy (J) Estimated Energy (J) % Error

News

Travel

Personal

2.983

2.219

0.735

2.800

2.200

0.730

6.0

5.0

0.5

Sujit Dey

Energy Aware Application OptimizationsEnergy Aware Application Optimizations

Start

BreakingNews

Politics Sports Enter. Finance

BNHeadlines

PoliticsHeadlines

CricketHeadline

s

BasketballHeadlines

Science

Fin.Headlines

Enter.Headlines

ScienceHeadlines

BNSummary

PoliticsFull Story

PoliticsSummary

ScienceSummary

Weather

ScienceFull Story

CricketSummary

CricketFull Story

BasketballSummary

BasketballFull Story

Enter.Full Story

Enter.Summary

Fin.Summary

Fin.Full Story

0.30.1 0.1 0.2 0.05

0.150.1

10.5

0.5 1 0.40.20.3

0.9

0.1

0.6

0.4

0.8

0.3

1.0

1.0

0.2 0.9

0.1 0.4

END

0.4

0.9

0.3

0.1

Fin.Headlines&Summary

0.5

BreakingNewswithSummary

Aggregation

Headlines

Deletion

Start &Weather

Migration

PoliticsHeadlines

PoliticsFull Story

Splitting

Sujit Dey

Exploration Methodology for Energy-Efficient Web Application Design

Exploration Methodology for Energy-Efficient Web Application Design

Probabilistic Web Flow Graph

TransactionGenerator

ClientEnergy Model

EnergyEstimator

Usage Profile Web Design

N1

N2 N3 N4

N6N5Access Probability

Web Page(b_tran, b_recv, priority)

Sujit Dey

Future EffortsFuture Efforts

The Internet

Base Station

Wireless GatewayServer

Web Servers

ConfigurableHandheld Client

BrowserApplication

ShaperApplication

Shaper Network ConditionsNetwork

Conditions

ApplicationRequirementsApplication

Requirements

Access Profile

Access Profile

Client Characteristics

Client Characteristics

WLANWCDMA

Sujit Dey

Network Aware Application Adaptation Network Aware Application Adaptation

• Factors influencing network environment– Different network technologies (WLAN,WCDMA,HDR)– Time varying bandwidth inside a network technology

• Example : Rate Adaptation in WCDMA/HDR

– User Mobility– Network Load

• Effect of the above factors on applications– Energy– Latency– Error Quality

• Objective: Adapt applications according to the current network environment to minimize energy consumption, latency of communication, and error quality.

Sujit Dey

Considering Network Effects on EnergyConsidering Network Effects on Energy

• Extend energy model to incorporate network effects– simulation based

• Network Parameters– Signal Strength

• determines “transmission power level”• no of re-transmissions

– Network Load• idle period

• Energy Model

Energy = TT(SS) * ( E(R,T,C,L,Enc,Comp) + TR(SS) ) + I(Network Load)

TT(SS) & TR(SS) : Table lookups based on signal strengthI (Network Load) : Table lookup based on network load

Sujit Dey

Test-bed EnvironmentTest-bed Environment

Needs Measurement (PalmVII,iPaq)

Simulation (OPNET)

Different access technologies

Realistic Network Environment

Ability to Instrument System Parameters

Accurate Measurement of Energy & Latency

Sujit Dey

ConclusionConclusion

• We developed an energy model and optimization techniques to adapt application to conserve energy

• We are currently working to develop network aware adaptation techniques

• We want to develop a test-bed environment to demonstrate and validate our ideas

Sujit Dey

Network-Aware Application AdaptationNetwork-Aware Application Adaptation

• Idea : To deliver information to users transparent of network conditions

• Means– Server : Re-organization of content

– Techniques : Filtering, Splitting, Clustering

» Step 1 : Filter out information which may not be useful

» Step 2 : Split Information into atoms of information

» Step 3 : Combine the atoms of information based on history information

– Issues :

» Temporary Storage of transformation

» Maintenance of connection state information

– Client : Pre-fetching & Presentation