2011 20 ppt

Data Dissemination in Opportunistic Networks

Radu-Ioan CiobanuCiprian Dobre

University Politehnica of BucharestFaculty of Automatic Control and Computers

Bucharest, Romania

May 25, 2011

IntroductionTaxonomy

AnalysisConclusions

Layout

1 Introduction

2 Taxonomy

3 Analysis

4 Conclusions

R.I. Ciobanu, C. Dobre Data Dissemination in Opportunistic Networks


AnalysisConclusions

Layout

1 Introduction

2 Taxonomy

3 Analysis

4 Conclusions



AnalysisConclusions

Introduction

Opportunistic networksnetworks composed of mobile nodes that may not have a directconnection between themroutes are build dynamically as topologies are unstablenodes act according to the store-carry-and-forward paradigmdata dissemination is usually based on a publish/subscribe model

Paper goalspresent the categories of a proposed taxonomy that captures thecapabilities of data dissemination techniques in opportunisticnetworkssurvey relevant data dissemination algorithms and analyze themusing the proposed taxonomy



AnalysisConclusions

Layout

1 Introduction

2 Taxonomy

3 Analysis

4 Conclusions



AnalysisConclusions

Taxonomy

DisseminationTechniques

DisseminationTechniques

NetworkOrganizationNetwork

OrganizationNode

CharacteristicsNode

CharacteristicsContent

CharacteristicsContent

CharacteristicsSocial

AwarenessSocial

Awareness

WithInfrastructure

WithInfrastructure

WithoutInfrastructureWithout

Infrastructure

NodeStateNodeState

NodeInteractionNode

Interaction

ContentAnalysisContentAnalysis

ContentOrganizationContent

Organization

SociallyUnawareSociallyUnaware

CommunityBased

CommunityBasedStatefulStateful

StatelessStateless

HybridHybrid

NodeDiscoveryNode

Discovery

ContentIdentificationContent

Identification

DataExchangeData

Exchange

CommunityDetection

CommunityDetection

CommunityStructure

CommunityStructure



AnalysisConclusions

Categories

Network organizationwith infrastructure

exploit nodes with high centrality (hubs) and build an overlaynetwork between them

without infrastructure

assume no infrastructure (costly to maintain and highly unstable)

Node characteristicsnode state

signifies if and how a node stores information about the nodesencountered so farthree approaches: stateful, stateless, hybrid

node interaction

must be efficient, because contact duration can be extremely lownode discovery - depends on the type of mobile devices being usedcontent identification - the way nodes represent the data internallyand how they “declare” itdata exchange - how nodes transfer data to and from each other



AnalysisConclusions

Categories (2)

Content characteristics

use case for opportunistic networks is the sharing of content availableon mobile users’ devicescontent producers and consumers might not be connectedcontent organization

the way content is organizeddissemination usually based on pub/sub, so content usually organizedin channels

content analysis

the way in which the algorithm analyzes a content object and decidesif it will fetch it or notadvanced methods assign priorities (utilities) to content objects,based on various parameters (hop count, subscribers, communities)



AnalysisConclusions

Categories (3)

Social awareness

nodes represented by devices that belong to humansuser movements are conditioned by social relationshipssocially-unaware

do not assume the existence of a social structure that governs themovement or interaction of nodessimple dissemination methods

community-based

assumes users are grouped into communities, based on socialrelationshipsmost used social model is the “caveman model”utilities are computed taking into account communitiessecurity improvement (implicit trust relationships)two aspects: community detection and community structure



AnalysisConclusions

Layout

1 Introduction

2 Taxonomy

3 Analysis

4 Conclusions



AnalysisConclusions

Dissemination Techniques

Socio-Aware Overlaya publish/subscribe data dissemination solution that uses an overlaycreated over detected communitiesoverlay composed of nodes with high closeness centrality values(hubs or brokers)subscriptions in a community are propagated towards brokers, whothen send them to other brokersdynamically performs decentralized community detectionnodes classified as community nodes, familiar strangers, strangers,friend nodes

Wireless Ad Hoc Podcastingpublish/subscribe approach, with content organized into channels,episodes, chunks and piecesqueries used when nodes are in range: random/new episodessolicitation strategies used when two nodes do not have content foreach other (Most Solicited, Least Solicited, Uniform, InverseProportional)



AnalysisConclusions

Dissemination Techniques (2)

DTN Pub/Sub Protocol (DPSP)

nodes split into sources, sinks, forwardersnodes exchange subscription lists when they meet and use a set offilters to build queues of bundles to forward to the peer (KnownSubscription, Hop Count, Duplicate)queues sorted by priorities using four heuristics (Short Delay, LongDelay, Subscription Hop Count, Popularity)

ContentPlace

publish/subscribe technique that exploits learned information aboutusers’ social relationships to decide where to place user datasocially-aware algorithm, based on the caveman modelutility function used to associate values to data objects (computed asweighted sum of one component for each community a node hasrelationships with)five policies for computing the utility function (Most FrequentlyVisited, Most Likely Next, Future, Present, Uniform Social)



AnalysisConclusions

Analysis of Four Dissemination Techniques

an approach without an infrastructure should be considered for amore general casea hybrid approach doesn’t suffer under frequent topology changesand doesn’t require large amount of control trafficdata exchange performed using data units as small as possiblepublish/subscribe pattern used for content organizationsocial-based dissemination algorithms have better resultsthere is no single best approach

Node Characteristics Content Characteristics Social AwarenessNode Interaction

Hybrid N/A

N/A N/A N/A

DPSP N/A N/A N/A N/A

ContentPlace Stateless Data objects Utility function N/A

DataDissemination

Technique

NetworkOrganization Node

StateContentAnalysis

ContentOrganization

CommunityDetection

CommunityStructureNode

DiscoveryContent

IdentificationData

Exchange

Socio-AwareOverlay

OverlayInfrastructure

Bluetoothand WiFi

Encounterednodes and

cont. duration

Subscriptionsand list ofcentralities

PublishSubscribe

Simple andK-clique

algorithms

Contactduration and

no. of contacts

Ad HocPodcasting

NoInfrastructure

Broadcastbeacons

Bloomfilter

hash index

Episodesor chunks

Solicitationstrategies

PublishSubscribe

NoInfrastructure

Subscriptionlists

Selectionof bundles

Filters andpriority

heuristics

PublishSubscribe

NoInfrastructure

Bluetoothand WiFi

Set of channelsthe node is

subscribed to

PublishSubscribe

Cavemanmodel



AnalysisConclusions

Layout

1 Introduction

2 Taxonomy

3 Analysis

4 Conclusions



AnalysisConclusions

Conclusions

Conclusions

proposed a taxonomy for data dissemination techniques inopportunistic networksanalyzed four relevant dissemination algorithms using the proposedtaxonomyconcluded that the future of opportunistic networking lies in thesocial property of mobile networks

Future work

propose and implement a socially-aware opportunistic mobile wirelesssolution for communication based on the conclusions of our analysis(Social Dissemination)



AnalysisConclusions

Thank You!


Documents

2011 20 ppt