abhishek presentation october 2013

BITS PilaniHyderabad Campus

A Survey of Social Based Routing in Delay Tolerant Networks: Positive and Negative Social Effects

Abhishek ThakurCSIS,

BITS-Pilani, Hyderabad Campus

11/04/2023 Slide 2 NetClique.in Internal Presentations BITS Pilani, Hyderabad Campus

For links that are intermittent or poor in quality, end-to-end connectivity and routing may not be guaranteed especially in Extreme Scenario.

Limits of MANets and Connected Networks

Prob success (iid fail prob pf) over k links:

For E2E delivery must have all links up

But, expected # of failed links is

kfsfs pkppp )1()(;1

fkp


Probability of Delivery


Optimizations on classic ad-hoc and delay-tolerant networking algorithms and began to examine factors such as security, reliability, verifiability

– Node mobility would be exploited to help deliver message (mobility-assisted or store-carry-and-forward)

Delay-Tolerant Networking Architecture– http://tools.ietf.org/html/rfc4838

Bundle Protocol Specification– http://tools.ietf.org/html/rfc5050

Why? What?

http://tools.ietf.org/html/rfc4838





Deep Space Communications– Beyond near-earth– Landers, Orbiters, Deep Space Probes

Sensor Networks– Terrestrial: Ocean or Land Based– Extra-terrestrial objects (on planets, etc)

High-Stress Physical Environments– Battlefield, Civil Emergency, Submarines

This and next few slides – ref Kevin Fall - 2008

What is Extreme?


• Large Delays

• Intermittent and Scheduled Links

• Bandwidth Asymmetry

• Limited Power

• Limited Emission Requirements (LPI/LPD)

• Heterogeneous Network Architectures

• Link Security Needs

• Very Large Scale (e.g. sensor nets)

Communications Challenges


Unusual Devices


Extreme systems

• Do not (won’t) run IP– High overhead, for global routing– Intra-Oceans: investigate routing– Space: very limited routing [e.g. rover to

lander]– Sensors: novel or simple routing, low

power

• Domain-specific features:– Naming, delivery abstraction, QoS

But we don’t want to scrap existing (Internet) software and experience

Heterogeneous Architectures

The NASA Deep Space Network (DSN)

– 3 70m-antenna array [USA, Spain, Australia]

Underwater Acoustic Modems– Bottom-to-top comm under 20kbps to

6Km

Low-Power CMOS Radios– Conventional and (hopefully) UWB

SINGCARS and EPLRS Military Radios


Goals of the DTN architecture:– Interoperability across network architectures– Reliability robust to link and node failure

Components:– Reliable Message Overlay with Routing– Interoperability Gateways– Flexible Naming Scheme– Per-hop Authentication with CoS

Delay-Tolerant Network Architecture


End-to-End Message Service: “Bundles”– “postal-like” delivery over regional transports– Optional class of service/notification

Key Idea: Custody Transfer– Custodian owns reliable-delivery guarantee– Bundles transferred between custodians toward

destination– Sender may free resources upon successful custody

transfer

Reliable Message Overlay


“Classic” Concepts (Internet):– Routing: selecting best next hop for every possible

destination– Forwarding: sending packet to best next hop

• Typically, “on demand” [statistical multiplexing]• Forwarders know a-priori next hop for every destination

DTN Concepts:– Routing: selecting best DTN next hop for destination– Forwarding: sending a bundle p2p when possible– Custody Transfer: reliable intra-DTN delivery (with storage)

Routing, Forwarding and Custody Transfer


Bundle Routing Example

A

B

B

Custody Tr

ansfer

Intermittent Links

DS

B

B

B

CT

End-to-end Acknowledgement(Sent using bundles, path omitted for clarity)

Contact Schedule

Aircraft

HUMMV

Schedule

Aircraft

HUMMV

CT


Is it possible that data can be delivered?– path between a source and a destination maybe

always won’t exist

Solution– Traditional protocols: Internet (RIP, OSPF); Ad hoc (DSR,

AODV) would fail– Formerly, mobility viewed as evil; Now, it’s perfect– Node mobility would be exploited to help deliver message

(mobility-assisted or store-carry-and-forward)

Mobility-assisted routing


Two categories– auxiliary nodes assisted (ANA) routing

• a set of special auxiliary nodes needed to assist data delivery• VANETs etc. • Throw-boxes / Ferry / Courier nodes / Autonomous Agents …

– independent mobile nodes (IMN) routing • there is no additional participants in the deployment area• message delivery achieved by node’s inherent movement • Proactive & reactive• Flooding vs. Heuristics based

Overview of Routing schemes


Pure Flooding

Probabilistic Flooding

Utility Based Flooding – goes into Heuristics

Multiple copies get created; Delivery reports used to clean them up

Epidemic: Flooding-based


tX(Y): time since X last saw Y

Indirect location information diffused with node mobility

smaller timer closer distance For most mobility models

Utility-based Routing

A

D

BtB(D) = 100

t(D) = 0

t(D) = 26

t(D) = 68

tA(D) = 138

t(D) = 218

Last encounter timers

D D

Utility UX(Y) = f(tX(Y))

Policy: forward to B if UB(D) > UA(D) + Uth

(A. Lindgren et al. ‘03)


Other knowledge-based routing– MaxProp: A variation of Dijkstra’s algorithm

• Link weight: an estimate of delivery likelihood between two nodes

– MobySpace: each node maintains a high-dimension Euclidean space

• Euclidean space: to describe mobility pattern of each node• Encounter occurred: handover message only if the encountered node

has more similar mobility pattern with the destination.


Performance metric– Message delivery ratio

• The fraction of generated messages that are correctly delivered to the final destination within a given time period

– Transmission delay• The time from a message is generated through it is received by

destination

– Number of transmissions (copies)• The number of message exchange occurred between two nodes

Routing objective


The performance of 2-hop scheme is close enough to multi-hop scheme (Burns et al, ’05)

Spray and wait scheme– 2-hop relay scheme

• “Spray” a number of copies to the network, then “wait” until one of relay nodes meets the desination

– Limited number of copies to L• Multi-path diversity to reduce delay• Achieves O(1) per node capacity

2-hop relay (multiple copies)


source starts with L copies

whenever a node with L > 1 copies finds a new node, it hands over half of the copies (L/2) that it carries; Until L = 1

Binary Tree-based Spraying

Src

C

B

Dst

D

EF

D

D

D

DL = 4

L = 2

L = 2

L = 1

L = 1

L = 1L = 1


Routing issue in DTN is challenge, attracting more attentionCategory of routing scheme

Have good scalability of DTN by exploring node mobilityFuture direction: develop more realistic networks; from military to public

application– Vehicle-based networks– Pocket-switched networks– Social networks– Wildlife tracking networks

Summary

ANA scheme IMN scheme

Knowledge-basedFlooding-based

Routing schemes


• DTN Intro

• Node Mobility vs. User’s Social Relations and Behavior

• Positive Social Characteristics

• Negative Social Characteristics

Brief Information From Paper


Community Detection• Modularity, Network of Communities

Information Propagation• Spread factor, Spreading Time• Connection Density, Influencing Factor, Critical Nodes

Recommendation System(s)• Online shopping / reviews, Twitter / RSS ranking etc.

Security and Privacy• Anti-Spam, limitations of blacklists, multiparty Authorization.• Privacy / Anonymity and re-identification algo for anonymized social

network

Social Network Analysis


Custody Transfers tied to contact graphs and duration of contact

Communality, Centrality and Similarity of nodes

Time slot based graph vs. Each edge records number, time and period of encounters

Contact graphs for DTN nodes and Social graphs for owners are loosely identical.

Question: How do the encounter tables get shared on large DTNs? Possibly scope to research further.

Social property if DTN : Graphs


Since DTN nodes are carried by people, they will tend to mimic communities. {impact of device sharing}

Sociological Centrality metrics will imply that such nodes are strong relay nodes

• Degree, Betweenness & Closeness to Destination

Similarity to Destination User location, Destination User Interests etc. implies that such relay nodes are likely to get near the destination. {challenge other than location how to capture and model similarity – data generated, apps used etc.}

Social property if DTN : Community, Centrality, Similarity


Similarity can infer Friendship: Long lasting regular contacts coupled with common interests in real world.

Rational Selfish behavior of social nodes implies that they want to send/receive DTN data not act as relay’s for other’s data

… Friendship and Selfishness


Label Routing Hui & CrowCroft – Pocket Switched nodes – IEEE PerCom 2007

• Community included as part of label• Nodes also share their affiliations and Groups i.e. Social

communitiesSimBet Routing Daly and Haar (ACM Symposium … 2007)

• Explores bridge nodes using betweenness, centrality and similarity.

• Using Betweenness and Similarity to destination, the Utility of node as next hop can be evaluated

• Scales by estimating centrality using only locally visible information – but can have –ve effects

Approaches used for Benefiting from Social Characteristics


Bubble Rap Forwarding Hui & other (MobiHoc 2008)• Community and centrality• Fast transfer towards destination community – bubble up

towards global centrality and bubble within community • Changeless to in Hierarchical communities or for

communities on periphery (non-central)Social Multicasting Gao & others (MobiHoc 2009)

• Centrality metric and community metric for relay• Single data or multi-data multicast• For Single data, assumed uniform destination distribution• For Muti-data – use gateway nodes for multiple communities

Approaches used for Benefiting from Social Characteristics…


Homophily Based data diffusion Zang & others (MobiHoc 2009)

• Prioritizes data propagation order (contact duration / storage space limited)

• Homophily => shared interest {challenge to identify / model this}

• Share most similar data items between friends and most different Data Items between strangers

Social Multicasting Bulut & others (GLOBECOM 2010)• Social Pressure Metric [weighted by frequent long lasting regular contacts]

• Data sharing / storage complexity is highUser Centric Disseminataion Gao & Cao (INFOCOMM 2011)

Approaches used for Benefiting from Social Characteristics…


Questions ?

Thanks


Backup


http://www.dtnrg.org

http://irtf.org/dtnrg

http://www.nasa.gov/mission_pages/station/research/experiments/DTN.html

– Disruption Tolerant Networking

Slide references– http://

bwrc.eecs.berkeley.edu/php/pubs/pubs.php/762/Delay%20Tolerant%20Architecture%20v1%20Kevin%20Fall.ppt

– http://www.networks.howard.edu/hguo/docs/prest05.ppt– http://netlab.cs.ucla.edu/wiki/files/relaycast_icnp08_final.ppt

Additional References

http://www.dtnrg.org/

http://www.dtnrg.org/






http://netlab.cs.ucla.edu/wiki/files/relaycast_icnp08_final.ppt



http://www.networks.howard.edu/hguo/docs/prest05.ppt





• Bundle, Payload, Fragment• Bundle node• Bundle protocol agent• Convergence layer adapters• Application agent• Bundle endpoint• Forwarding, Registration, Delivery• Deliverability/Abandonment • Deletion, Discard• Transmission• Custody

Terminology

abhishek presentation october 2013

Technology