49
1 Routing in Disruption- Tolerant Networks Katia Obraczka University of California, Santa Cruz [email protected] http://inrg.cse.ucsc.edu/

Routing in Disruption-Tolerant Networks

  • Upload
    nat

  • View
    78

  • Download
    0

Embed Size (px)

DESCRIPTION

Routing in Disruption-Tolerant Networks. Katia Obraczka University of California, Santa Cruz [email protected] http://inrg.cse.ucsc.edu/. What are Disruption-Tolerant Networks?. Disruption-tolerant networks or DTNs. A.k.a, Delay-tolerant, Episodically-connected, - PowerPoint PPT Presentation

Citation preview

Page 1: Routing in Disruption-Tolerant Networks

1

Routing in Disruption-Tolerant Networks

Katia ObraczkaUniversity of California, Santa Cruz

[email protected]

http://inrg.cse.ucsc.edu/

Page 2: Routing in Disruption-Tolerant Networks

2

What are Disruption-Tolerant Networks?

• Disruption-tolerant networks or DTNs.

• A.k.a, – Delay-tolerant, – Episodically-connected,– Intermittently-connected networks.

Networks where end-to-end connectivity is NOT guaranteed.

Page 3: Routing in Disruption-Tolerant Networks

3

DTN Applications: Military Operations

Page 4: Routing in Disruption-Tolerant Networks

4

DTN Applications: Environmental Monitoring

CARNIVORES Project at UCSC: Monitoringcoyotes in the SantaCruz mountains.

Page 5: Routing in Disruption-Tolerant Networks

5

DTN Applications: Connecting Remote Communities

• Rural “kiosks”:– Shared among locals.– Selling/buying

agricultural products.– Banking and other

transactions.

Page 6: Routing in Disruption-Tolerant Networks

6

S

DX Xpath

disruption!

DTNs

Connectivity disruption due to:• Wireless propagation (fading, shadowing, etc.)• Duty cycling for power conservation.• Mobility.

Regular Network DTN

S D

End-to-end path

S

DX

Xpath disruption!

nodelink

Page 7: Routing in Disruption-Tolerant Networks

7

What’s the big deal?• Routing protocols have always assumed

end-to-end connectivity.– Table-driven (proactive) protocols (e.g., Internet

routing) can recover from infrequent topology changes.

– On-demand (reactive) protocols (e.g., MANET routing) can recover from frequent but short-lived outages.

• But what if “outages” are frequent and long-lived?– “Traditional” routing simply drops packets!

Page 8: Routing in Disruption-Tolerant Networks

8

DTN’s Routing Paradigm Shift

• Before DTNs:– Space dependency.– Network routing:

given graph G(V,E), find shortest path between source-destination.

– Store-and-forward routing.

• After DTNs:– Space and time

dependency.– Network as a time-

varying graph G(V,E(t)).

– Links are a function of time.

– Links as “contacts”.– Store-carry-and-

forward routing.

Page 9: Routing in Disruption-Tolerant Networks

9

Types of Contacts• Scheduled contacts

– E.g. satellite links, message ferry.

– All info known.

• Probabilistic contacts– Statistics about

contacts known.– E.g., bus, sensors with

random wake-up schedule.

• Opportunistic contacts– Not known a priori.– E.g., tourist car that

happens to drive by.

Page 10: Routing in Disruption-Tolerant Networks

10

Designing DTN Routing/Forwarding Protocols

• What information is available?– Oracles.– Contacts, contact statistics, queuing, traffic, buffer

capacity, etc.– None.

• How much information is known?– No knowledge.– Partial knowledge.– Complete knowledge.

• Trade-offs?

Page 11: Routing in Disruption-Tolerant Networks

11

Routing/Forwarding under Intermittent Connectivity

Scheduled/, (partially) known contacts (e.g., buses).

Enforced contacts with specialized nodes (e.g., ferries).

What about unknown contacts?– Contacts not known in advance.– No specialized nodes. i.e., only mobility of the nodes

themselves is available.

Opportunistic (mobility-assisted) routing

Page 12: Routing in Disruption-Tolerant Networks

12

Our Focus: Opportunistic Routing

Generality and simplicity: no knowledge assumed.

• Opportunistic routing paradigm:– At every hop, node decides whether to:

• Forward and/or• Store-and-carry.

– Store-carry-and-forward.

Page 13: Routing in Disruption-Tolerant Networks

13

Current Research

• Two thrusts:– Utility-based controlled replication in

heterogeneous DTNs.

– Steward-Assisted Routing (StAR)• Routing framework that works well in both

connected networks and networks prone to frequent, long-lived disconnections.

Page 14: Routing in Disruption-Tolerant Networks

14

Utility-Based Replication in Heterogeneous DTNs

• In collaboration with Akis Spyropoulos and Thierry Turletti, INRIA Sophia-Antipolis.

T. Spyropoulos, T. Turletti, and K. Obraczka, ``Utility-based Message Replication for Intermittently Connected Heterogeneous Networks'', in the Proceedings of the 1st. IEEE WoWMoM Workshop on Autonomic and Opportunistic Communications (AOC) 2007.

T. Spyropoulos, T. Turletti, and K. Obraczka "Routing in Delay-Tolerant Networks Comprising Heterogeneous Node Populations”, in preparation.

Page 15: Routing in Disruption-Tolerant Networks

15

Opportunistic Routing• Epidemic Routing: everyone gets a copy (Vahdat et al.

‘00).

(+) Fast (for low traffic loads!). (-) Too many resources!

• Controlled Replication: give a copy to only L nodes (Small et al. ’05, Spyropoulos et al. ’05).

(+) Control transmissions/energy/buffer space per message!

(+) few copies (<10% of nodes) give good performance (not always!)

(-) Replication is greedy: blindly choose relays!

Page 16: Routing in Disruption-Tolerant Networks

16

Greedy Replication: A Good Scenario

Covered by Relay 1

3

S

D

1

2

4

5

6

7

8

9

10

11

12 13

14

1615

Covered by relay 2

2 copies

Relays are highly mobile

Relays routes are uncorrelated

Nodes have homogeneous behavior

Page 17: Routing in Disruption-Tolerant Networks

17

Heterogeneous Networks

Community (local) Nodes

Base Stations or Static Sensors

Fast/Mobile Nodes

stay inside community

Roam around network(infrequent)

Page 18: Routing in Disruption-Tolerant Networks

18

Greedy Replication in Heterogeneous Environments

• If p% of the nodes are “useless” (local):– Delay increase ≈ c1 + c2/(1-p) (c1, c2

constants, and c1 + c2 = 1) compared to homogeneous case.

Src

Dst

3 copies

Move “locally”

Roaming NodeGood Relay!No more copies!

Page 19: Routing in Disruption-Tolerant Networks

19

Goal: Discover “Best” RelaysHow? Maintain some utility function.

– What does “better” mean?

• 2 types of utility function: Destination-Dependent (DD) Utility:

“Node X is a good next hop for destination D”.– E.g., friendship with D, proximity to D, etc. Destination-Independent (DI) Utility:

“Node X is a good next hop for all destinations”.– E.g., moves more frequently/faster around the network,

is a “social hub”, has higher resources (e.g. vehicle with no energy limitations), etc.

Page 20: Routing in Disruption-Tolerant Networks

20

What to Do With Utility?Approach: Give L copies to best relays:

(+) Control resources: IMPORTANT! (e.g., scarce resources like battery power)

Page 21: Routing in Disruption-Tolerant Networks

21

Utility-based (“Smart”) Replication

• Message from S to D.

• S starts with 1 message copy and L “fwd tokens”.

• UX(Y): Utility of node X for destination Y.

SD

D

Option 1) If UB(D) > UA(D) (relative utility)

Option 2) If UB(D) > Uth (absolute utility)

L tokens

A

D

n>1 tokens

B

D

n/2n/2

Page 22: Routing in Disruption-Tolerant Networks

22

Last Seen First (LSF) Replication

• Age-of-last-encounter timer: “I last saw node D less than 10 minutes ago”.

• tX(Y): last time X saw Y.

• UX(Y) = 1/(tX(Y)+1)– If my timer is smaller, then I’m a better relay

for that destination.– DD utility.

Page 23: Routing in Disruption-Tolerant Networks

23

LSF Spraying

0

0.5

1

1.5

2

2.5

3

3.5

4

L=10, K=40 L=16, K=40 L=10, K=50 L=16, K=50

Number of copies, Tx Range

Del

ay (

SW

) / D

elay

(L

SF

) LSF (transitivity)

LSF (no transitivity)

Simulation Results (LSF)

Community-based Mobility (Infocom’07)Four types of nodes (100 total)

Community nodes (40%), Local nodes (40%) Roaming nodes (10%), Static nodes (10%)

more copies => less costly mistakes

Page 24: Routing in Disruption-Tolerant Networks

24

Most Mobile First (MMF) Replication

• Utility of X = Label of node X.• DI.

• Preference Order (): Label1 Label2… LabelN.

– E.g. based on statistical properties or characteristics. – Order may also depend on destination’s label.

Community

Base Station

Pedestrian

Pedestrian

VehicleVehicle

Sensor

Page 25: Routing in Disruption-Tolerant Networks

25

Simulation Results – MMF (1)

• Labels: “roam” “community” ”local” “static”• MMF1: give only to {“roam” || “community”}• MMF2: {“roam” || “community”} && {Prob{roam} > 0.15}

Same scenario as before

4 types of nodes

MMF spraying (Scenario 1)

0

0.5

1

1.5

2

2.5

3

3.5

MMF1(L=10) MMF2(L=10) MMF1(L=16) MMF2(L=16)

Protocol (# of copies)

Del

ay (

SW

) / D

elay

(M

MF

) sparse (8% connectivity)

almost connected (52% connectivity)

Page 26: Routing in Disruption-Tolerant Networks

26

Simulation Results – MMF (2)• 2 types of nodes:

“mobile” and “static”

• Algorithm: give only to “mobile”

• Nodes = 100

MMF Spraying (Scenario 2)

1

1.5

2

2.5

3

3.5

4

4.5

5

5.5

0.4 0.3 0.2 0.1 0.05percentage of mobile nodes (p)

De

lay

(S

W)

/ D

ela

y (

MM

F) K = 40, L = 16

K = 50, L = 10K = 50, L = 16K = 60, L = 16

many “good” options => fewer mistakes by “greedy”

few “good” options => mistakes cost!

Page 27: Routing in Disruption-Tolerant Networks

27

Most Social First (MSF) Algorithm

• Mobility statistics not always available or dynamic (e.g. node in car, then in office)

• Estimate node “sociability” online– tn = [(n-1)T, nT] (nth time window – duration T)

– Ni(n): set of nodes seen by node i during tn

– Si(n) = |Ni(n)|/T: sociability of i during tn

– Running average:

– Utility of node i: Ui =

(n)aSSa)(1S iii ˆˆ

iS

Page 28: Routing in Disruption-Tolerant Networks

28

Tuning MSF’s Parameters• Can past predict future?

• How do we set T (window) and a (weight of new sample)?

– Depends on (i) mobility/interaction patterns, (ii) horizon of prediction

– time-homogeneous: past reliable predictor!– periodic in T*: set T to T*

– shorter scale prediction: smaller T, larger a, higher moments

• Number of copies: find lower bound given target delivery ratio and message TTL.

Page 29: Routing in Disruption-Tolerant Networks

29

Simulation Results – MSF

2 types of nodes: “mobile”, “static” MMF: knows “mobile” labels beforehand MSF: identifies “mobile” nodes online. Same performance!

T = 1000, a = 0.8, L = 10

MSF Spraying (Delivery Ratio)

00.10.20.30.40.50.60.70.80.9

1

0.1 0.2 0.3

Ratio of "useful" nodes (p)

de

live

ry p

rob

ab

ility

MSF Spraying (Delivery Delay)

0

1000

2000

3000

4000

5000

6000

7000

0.1 0.2 0.3Ratio of "useful" nodes (p)

de

liv. d

ela

y(t

ime

un

its

)

Greedy

MMF

MSF

Page 30: Routing in Disruption-Tolerant Networks

30

ConclusionUtility-based vs. Greedy Replication in

Heterogeneous Networks: up to 4-5x improvement– Few “good” options => bigger gains.– Small budget of copies => bigger gains.

MSF: Generic, adaptive algorithm to discover “social” nodes.

Can be a building block for more complex schemes.

Page 31: Routing in Disruption-Tolerant Networks

31

Ongoing and Future Directions

• Modeling encounter-based (epidemic) protocols in heterogeneous environments.– Multiple classes of nodes. – Different mixing characteristics.– Fluid model approximation.

Page 32: Routing in Disruption-Tolerant Networks

32

Ongoing and Future Directions (Cont’d)

• Trace-based evaluation of Smart Replication. – E.g., National University of Singapore class

schedules.

• Hybrid DI/DD utility functions; more sophisticated sociability estimators.

Page 33: Routing in Disruption-Tolerant Networks

33

Steward-Assisted Routing (StAR)

• Routing framework that performs well in both connected networks as well as networks prone to frequent, long-lived disconnections.

Page 34: Routing in Disruption-Tolerant Networks

34

StAR

• In collaboration with Jay Boice (UCSC MSc, May 2007) and J.J. Garcia-Luna.

• J. Boice, J.J. Garcia-Luna Aceves, K. Obraczka, ``Disruption-Tolerant Routing with Scoped Propagation of Control Information'', in Proceedings of the IEEE International Conference on Communications (ICC) 2007.

• J. Boice, J.J. Garcia-Luna Aceves, and K. Obraczka, ``On-Demand Routing in Disrupted Environments'', BEST PAPER AWARD, in Proceedings of the IFIP/TC6 Networking 2007.

• J.J Garcia-Luna Aceves, K. Obraczka, and J. Boice, “An On-Demand Routing Framework for Disruption-Tolerant Networks”, under submission.

Page 35: Routing in Disruption-Tolerant Networks

35

StAR Highlights

• Combines on-demand (intra-partition) with opportunistic (inter-partition) routing.– Use of relays, or stewards, to deliver data to

partitioned destinations.

• No a-priori topological knowledge.– Use past connectivity information to predict future

communication opportunities.

• Scopes temporal and spatial dissemination of routing information.

Page 36: Routing in Disruption-Tolerant Networks

36

SCIP

• Scoped Contact and Interest Propagation.

• Limits scope of routing information.– Nodes get routing info

for destinations of interest.

– Nodes only keep info for d if they are on the path from s to d.

Example: s1 and s2 interested in d.

Page 37: Routing in Disruption-Tolerant Networks

37

Steward Selection

• Steward selected for given destination.

• Use sequence numbers and number of hops to select local steward for destination d.– Steward has most

recent sequence number.

– If sequence numbers are equal, choose node with lowest number of hops to destination.

Example: one steward per destinationper partition.

Page 38: Routing in Disruption-Tolerant Networks

38

Well-Connected Topologies

• 100 nodes in 3600x500m area with full connectivity.

• Static and random waypoint mobility.• Comparison against AODV and OLSR.

StAR performs well with full connectivity and under short-lived disconnections.

Page 39: Routing in Disruption-Tolerant Networks

39

As Connectivity Decreases…

Decreasing connectivity

PDR

AODV

Epidemic

StAR

Gridded mobility.

Decrease connectivity byincreasing grid dimension.

Page 40: Routing in Disruption-Tolerant Networks

40

StAR Experiments: Special Operation Scenarios

• Real experiments with StAR on testbed with static nodes and mobile robots.

• Collaboration with Prof. Weitzenfeld’s robotics lab at ITAM, Mexico.

K. Obraczka, J. Boice, L. Martínez-Gómez, J. P. Francois, A. Levin-Pick, A. Weitzenfeld, “StAR: Ad-Hoc Wireless Networking for Autonomous Multi-Robot Coordination”, to appear in the Proceedings of the 1st. IEEE/ACM Robocom, October 2007.

Page 41: Routing in Disruption-Tolerant Networks

41

StAR Testbed

Eagle Knights modified robot with local camera and 802.11 communicationcapabilities.

The original robot architecture is extended with: (1) the Crossbow Stargatemanaging wireless communication and local vision and (2) a webcam for sensing.

Page 42: Routing in Disruption-Tolerant Networks

42

Results

Topology 2: Static sensors with mobile intermediate node. Sensor 4 sends images to sink node 7 through intermediate mobile node 2 and static node 1.

TABLE IIPerformance of AODV and StAR in Topoloy 2

Image Deliveries DeliveryRatioAODV 48 51.11%StAR 90 100.00%

Page 43: Routing in Disruption-Tolerant Networks

43

Summary

• StAR as routing framework that operates well in both well-connected networks as well as networks prone to episodic connectivity.

• No a-priori knowledge, e.g., node schedules, location, etc.

• Combines on-demand (intra-partition) with opportunistic (inter-partition) routing.– Use of relays, or stewards, to deliver data to

partitioned destinations.

Page 44: Routing in Disruption-Tolerant Networks

44

Future Work

• Use other sources of information to improve performance.– Full/partial node schedules, GPS, etc.

• Investigate other metrics for steward selection.

• Explore different message replication strategies.

• Integrate with work on utility-based opportunistic routing.

Page 45: Routing in Disruption-Tolerant Networks

45

Summary of Other Activities

Page 46: Routing in Disruption-Tolerant Networks

46

Energy-Efficient Medium Access (Task 4)

• Novel efficient and flexible medium access framework form MANETs.

• With J.J. Garcia-Luna and Venkatesh Rajendran (UCSC PHD, May 2007).

V. Rajendran, K. Obraczka and J.J. Garcia-Luna, ``A DYNAmic Multi-channel Medium Access Framework for Wireless Ad Hoc Networks'', BEST PAPER AWARD in the Proceedings of the 4th. IEEE International Conference on Mobile Ad-hoc and Sensor Systems (MASS) 2007.

V. Rajendran, Medium Access Control Protocols for MANETs, PhD Dissertation, UCSC, 2007.

V. Rajendran, K. Obraczka and J.J. Garcia-Luna, Application-Aware Medium Access for Sensor Networks, 2nd IEEE International Conference on Mobile Ad-hoc and Sensor Systems (MASS), November 2005.

Page 47: Routing in Disruption-Tolerant Networks

47

Energy Consumption Modeling and Prediction (Task 4)

• Models for energy consumption and network lifetime prediction.

• Collaboration with Prof. R. Manduchi, UCSC CE.

C. Margi, Energy Consumption Trade-Offs in Power-Constrained Networks, PhD Dissertation, UCSC, 2006.

C. Margi, K. Obraczka, R. Manduchi, Characterizing System Level Energy Consumption in Mobile Computing Platforms, IEEE WirelessCom 2005, June 13-16, 2005

C. Margi, V. Petkov, K. Obraczka, R. Manduchi Characterizing Energy Consumption in a Visual Sensor Network Testbed, IEEE/Create-Net TridentCom 2006, March 1-3, 2006.

Energy Consumption Trade-offs in Visual Sensor Networks”. C. B. Margi, R. Manduchi , K. Obraczka. SBRC 2006, May 29 - June 02, 2006.

Page 48: Routing in Disruption-Tolerant Networks

48

Mobility Models for Wireless Networks (Task 1)

• First Statistical-Equivalent Model (SEM) to characterize random waypoint mobility.

• Collaboration with Profs. B. Sanso and A. Kottas, UCSC Applied Math, and K. Viswanath (NTT Labs).

K. Viswanath, A. Kottas, B. Sanso, and K. Obraczka, ``Statistical Equivalent Models for Computer Simulators'', in press to appear in the special issue of Simulation: Transactions of the Society for Modeling and Simulation International on ''Advances in Performance Evaluation of Computer and Telecommunication Systems”.

K. Viswanath, K. Obraczka, A. Kottas, B. Sanso, A Statistical Equivalent Model for Random Waypoint Mobility: A Case Study, IEEE SMC SPECTS 2006.

K. Viswanath and K. Obraczka, Modeling the Performance of Flooding in MANETs (Extended Version), Computer Communications Journal (CCJ) 2005.

Page 49: Routing in Disruption-Tolerant Networks

49

Robust Routing for Network Fault-Tolerance and Security (Task 6)

• Novel game-theoretic stochastic routing framework as proactive alternative to today's reactive approaches to route repair.

• Collaboration with Prof. J. Hespanha, affiliated with UCSB’s ICB (Army UARC) and Prof. S. Bohacek at UDel, funded by the Communications&Networking Army CTA.

S. Bohacek, J.P. Hespanha, C. Lim, and K. Obraczka, ``Game Theoretic Stochastic Routing for Fault Tolerance on Computer Networks'', in press to appear in the IEEE Transactions Parallel and Distributed Systems.

C. Lim, Scalable Multi-path Routing for Robust Communication, PhD Dissertation, USC, 2006.

G. Huang, Robust and Secure Routing in MANETs, MSc Theis, UCSC, 2006. C. Lim, S. Bohacek, J. Hespanha and K. Obraczka, Hierarchical Max-Flow

Routing, IEEE Globecom 2005. S. Bohacek, J. Hespanha, J. Lee, C. Lim and K. Obraczka, A New TCP for

Persistent Packet Reordering, IEEE/ACM Transactions on Networking, Vol. 14, No.2, April 2006.

R. Guru, G. Huang and K. Obraczka, An Integrated and Flexible Approach to Robust and Secure Routing in MANETs, IEEE IC3N, August 2005.