25
Distance and Probabilistic Routing Protocol using History of Encounters and Transitivity (DPRoPHET) in Delay Tolerant Network (DTN) Advisor Professor: Keecheon Kim Phearin Sok MBC Laboratory Computer, Information, and Communication Engineering Konkuk University Published in International Conference on Applied Materials and Electronics Engineering (AMEE) 2013, Hong Kong Friday, May 31, 2013

DPRoPHET in Delay Tolerant Network

Embed Size (px)

DESCRIPTION

PRoPHET uses its delivery predictability of node encounters and transitivity to forward bundles to its neighbor node. Regardless of their distance, it faces delivery dilemmas in a source node and drawbacks of low delivery ratio and high delay in case two or more neighbor nodes carry equal delivery predictability. To solve such consequences, we propose a Distance and Probabilistic Routing Protocol using History of Encounters and Transitivity (DPRoPHET) with the use of cross layer implementation for distance value retrieval. Our simulation results show that, by adding distance metric to the existing delivery predictability vector, DPRoPHET outperforms PRoPHET.

Citation preview

Page 1: DPRoPHET in Delay Tolerant Network

Distance and Probabilistic Routing Protocol using History of Encounters

and Transitivity (DPRoPHET) in Delay Tolerant Network (DTN)

Advisor Professor: Keecheon Kim

Phearin Sok

MBC Laboratory

Computer, Information, and Communication Engineering

Konkuk University

Published in International Conference on Applied Materials and Electronics Engineering (AMEE) 2013, Hong Kong

Friday, May 31, 2013

Page 2: DPRoPHET in Delay Tolerant Network

Contents• Background

PRoPHET Overview

• Problems Statement

• Solution: DPRoPHET Routing Protocol in DTN

Distance Mechanism

Cross Layer Model

DPRoPHET Delivery Probability

DPRoPHET Scenarios

• Simulation and Evaluation

• Conclusion

• Key References

Page 3: DPRoPHET in Delay Tolerant Network

Background

Page 4: DPRoPHET in Delay Tolerant Network

PRoPHET in DTN• Probabilistic Routing Protocol using History of Encounters and

Transitivity

• All data forwarding happens at opportunistic encounters

• Using probabilistic metric called Delivery Predictability (DP)

a: DP generator Node

b: Destination Node

A

B

C

S

Note: P(a, b) > P(c, b) : Bundles for destination B

are likely to be forwarded to A rather than C

P(a, b)∈[0,1]

Page 5: DPRoPHET in Delay Tolerant Network

The Three DP Calculations

• Transitivity Equation

initoldbaoldbaba PPPP )1( ),(),(),(

oldbaba PP ),(),(

βPPPPP (b,c)(a,b)old(a,c)old(a,c)(a,c) )1(

A B

S

A

C

B

S

Pinit∈[0,1]: Initialization constant

γ∈[0,1): Aging constant

κ: Number of time units

β∈[0,1]: Impact scaling constant

• Direct Encounter – Main Equation

• Aging Equation

Page 6: DPRoPHET in Delay Tolerant Network

PRoPHET Principal Phases

• New Neighbor Awareness

Done by lower layers

PRoPHET is notified if neighbor nodes are in range

• Two-part Information Exchange Phase

1st part: exchange summary vector (Hello procedure)

o Delivery Predictability Values

o Bundle Information, Buffer Size…

2nd part: forward the bundles

Page 7: DPRoPHET in Delay Tolerant Network

Problems Statement

Page 8: DPRoPHET in Delay Tolerant Network

PRoPHET Drawbacks

• Lack of distance awareness

Impact: Introduce delay

• More resource consumption for equal DP value

Impact: Storage shortage, Bundles dropping

• Highly fluctuating DP value

Impact: Fault forwarding decision

• No Acknowledgement

Impact: Stale copies of delivered message

Page 9: DPRoPHET in Delay Tolerant Network

Solution

Page 10: DPRoPHET in Delay Tolerant Network

DPRoPHET Routing Protocol

• Modified from PRoPHET protocol based on Neighbor

Node Distances

Therefore, Neighbor Node Distance can be

retrieved:4

22

Pr L

hhGGPd rtrtt

Ld

hhGGP rtrtt

4

22

Pr

• Received Power Strength of Two-ray Ground

Reflection Model

Page 11: DPRoPHET in Delay Tolerant Network

Cross Layer Model

Application

Bundle (DTN)

Transport

Network

Link

Physical

Sh

are Reg

istry

DTN Cross Layer Model

4

22

Pr L

hhGGPd rtrtt

initoldbaoldbaba PPPP )1( ),(),(),(

Page 12: DPRoPHET in Delay Tolerant Network

New Delivery Predictability

‽ Which message is suitable for distance retrieval?

Hello message to exchange DP value during the encounter

d

initold(a,b)old(a,b)(a,b) εPPPP 1

ε∈[0,1): Distance scaling constant

d: Distance between nodes from the Share Registry

Note: Larger ε, higher impact on the delivery predictability

• Direct Encounter Calculation for DPRoPHET

Page 13: DPRoPHET in Delay Tolerant Network

• Transitivity Equation

oldbaba PP ),(),(

βPPPPP (b,c)(a,b)old(a,c)old(a,c)(a,c) )1(

• Direct Encounter – Main Equation

• Aging Equation

DPRoPHET Delivery Predictability

d

initold(a,b)old(a,b)(a,b) εPPPP 1

Maintained from

PRoPHET

Page 14: DPRoPHET in Delay Tolerant Network

DPRoPHET Scenarios

Equal Delivery Predictability

A

B

C

D

E

F

SP(b,d)=0.6

P(a,d)=0.6

Near-equal Delivery Predictability

A

B

C

D

E

F

S

P(a,d)=0.605

P(b,d)=0.6

Page 15: DPRoPHET in Delay Tolerant Network

Simulation and Evaluation

Page 16: DPRoPHET in Delay Tolerant Network

Simulation

• Simulation Setup

Parameter Value

NS2 Version 2.35

Mobility Model Random Waypoint (RWP)

Area 2000m x 2000m

Number of nodes 40

Speed 2.5 – 20 m/s

Radio Propagation

ModelTwo-ray ground reflection

Message size 100 KB

Buffer size Up to 100 MB

Bundle lifetime 750 seconds

Transmission Timeout 1000 seconds

Simulation time 5000 seconds

Constants Value

Pinit 0.75

β 0.25

γ 0.98

ε 0.3

Simulation Setting

Protocol Constants Setting

• Languages:

C++

OTcl

Page 17: DPRoPHET in Delay Tolerant Network

DPRoPHET Implementation

• Node Distance double TwoRayGround::getDist(double Pr, double Pt, double

Gt, double Gr, double hr, double ht, double , double )

{ /* Get quartic root */

return sqrt(sqrt(Pt * Gt * Gr * (hr * hr * ht * ht) / Pr));

}

Double TwoRayGround::Pr(PacketStamp *t, PacketStamp *r, WirelessPhy *ifp){

……

if(loc_d <= crossover_dist) {

Pr = Friis(t->getTxPr(), Gt, Gr, lambda, L, d);

dist = getDist(Pr,t->getTxPr(),Gt,Gr,ht,hr,0.0,0.0);

r->getNode()->add_neigh_dist(t->getNode()->nodeid(), (double)dist,

Scheduler::instance().clock());

return Pr;

}

else{

Pr = TwoRay(t->getTxPr(), Gt, Gr, ht, hr, L, d);

dist = getDist(Pr,t->getTxPr(),Gt,Gr,ht,hr,0.0,0.0);

r->getNode()->add_neigh_dist(t->getNode()->nodeid(), (double)dist,

Scheduler::instance().clock());

return Pr;

}

}

Page 18: DPRoPHET in Delay Tolerant Network

DPRoPHET Implementation• Protocol Implementation void BundleAgent::recv(Packet* pkt, Handler*) {

hdr_ip* iph=hdr_ip::access(pkt);

hdr_bundle* bh=hdr_bundle::access(pkt);

if (((u_int32_t)iph->daddr()==IP_BROADCAST) && (bh->type ==3)){

// Hello Message & Update neighbor information

int found=0, i=0;

while ((found == 0)&&(i < neighbors)) {

if (neighborId[i] == iph->saddr())

found=1;

else

i++;

}

If(found==0){

…… //initialize new DP

}

……

dist = dist_table[neighborId[i]];

ownDP[i]=ownDP[i] + (1-ownDP[i]) * 0.75 + pow(0.3, dist);

……

}//end Hello Message (type=3)

……

}

Page 19: DPRoPHET in Delay Tolerant Network

Evaluation

• Message Delivery Ratio Improves the performance

Produces a higher ratio

Average Increment: 5%

35%40%

60%

70%

83%

33%

42%

65%

77%

92%

0%

20%

40%

60%

80%

100%

20 40 60 80 100

Message D

eliv

ery

Ratio

Buffer Size (MB)

Message Delivery Ratio vs. Buffer Size

PRoPHET DPRoPHET

Page 20: DPRoPHET in Delay Tolerant Network

Evaluation

• Average Delay

220

330

430

455480

215

310

400420 435

0

100

200

300

400

500

20 40 60 80 100

Avera

ge D

ela

y (s

)

Buffer Size (MB)

Average Delay vs. Buffer Size

PRoPHET DPRoPHET

Reduce the delay of bundles delivery

Average Reduction: 30s

Page 21: DPRoPHET in Delay Tolerant Network

Evaluation

• Message Overhead

11500

22000

31000

36500

48000

11000

20000

28000

33000

44000

0

10000

20000

30000

40000

50000

20 40 60 80 100

Avera

ge F

orw

ard

ed M

essage

Buffer Size (MB)

Average Forwarded Messages vs. Buffer Size

PRoPHET DPRoPHET

Reduce the overhead of message exchange

Average Reduction: 2600 messages

Page 22: DPRoPHET in Delay Tolerant Network

Conclusion

Page 23: DPRoPHET in Delay Tolerant Network

To Conclude…

• A popular routing protocol in DTN – PRoPHET routing protocol and its drawbacks

• DPRoPHET considering the distance metric as an addition operant to produce better delivery predictability

• DPRoPHET produces equal or better performancesthan PRoPHET, in terms of

Increasing Delivery Ratio

Reducing Delivery Delay

Less Protocol Exchange Overhead

Page 24: DPRoPHET in Delay Tolerant Network

Key References

• Sok, P., Lee, S., & Kim, K.C. (2013). “DPRoPHET in Delay Tolerant Network.” 2nd International Conference on Applied Materials and Electronics Engineering (AMEE2013). Advanced Materials Research Vol. 684 (2013) pp 543-546

• Lindgren, A. et al., "Probabilistic Routing Protocol for Intermittently Connected Networks," IETF 6693, August 2012

• Lindgren, A., Doria, A., & Schelén, O. (2003). “Probabilistic routing in intermittently connected networks.” ACM SIGMOBILE Mobile Computing and Communications Review, 7(3), 19-20.

• …

Page 25: DPRoPHET in Delay Tolerant Network

Thank You…