Weak State Routing for Large Scale Dynamic Networks

Utku Günay Acer, Shivkumar Kalyanaraman, Alhussein A. Abouzeid

Rensselaer Polytechnic InstituteDepartment of Electrical, Computer & Systems Engineering

Which area should we NOT be working on in MOBICOM anymore?

Ans: Routing !

- Victor Bahl, Mobicom 2007 panel

Routing in Large-scale Dynamic Networks

Routing table entries: “state” = indirections from persistent names (ID) to locators

Due to dynamism, such indirections break

Problematic on two dimensions Dynamism/mobility =>

frequent update of state Dynamism + large scale

=> very high overhead, hard to maintain structure

We propose constructing routing table indirections using probabilistic and more stable state: WEAK STATE

Number of NodesN

ode M

obili

ty

A new class of State

Strong StateDeterministicRequires control

traffic to refreshRapidly

invalidated in dynamic environments

Weak StateProbabilistic

hintsUpdated locally

Exhibits persistence

STATEB

STATEB

Hard, Soft and Weak State

a b

INSTALL

STATEA

STATEA

REMOVE

Hard StateSoft State

UPDATE

Time elapsed since state

installed/refreshedWeak StateConfidence in

state information

Weak State is natural generalization of soft state

Random Directional Walks

Both used to announce location information (“put”) and forward packets (“get”)

Outline

Our Weak State RealizationDisseminating Information and

Forwarding PacketsSimulation ResultsDiscussion & Conclusion

An Instance of Weak State

The uncertainty in the mappings is captured by locally weakening/decaying the state

Other realizations are possibleProphet, EDBF etc…

{a,b,c,d,e,f}

Probabilistic in terms of membershi

p

SetofIDsGeoRegion

Probabilistic in terms of

scope

Consider a node aWhere is node a?

(i): It is in region ABwith probability 1

(ii) It is in region B with probability 2

(1 · 2)

Example: Weak State

128.113.

128.113.

128.113.62.

128.113.62.

128.113.50.

128.113.50.

xn

1x

2x

x

21

x

How to “Weaken” State?

Larger Geo-Region

Aggregation

SetofIDs -> GeoRegion

Aggregation: setofIDs

setofIDs: We use a Bloom filter, denoted by B.

m1 m2

….0 1 0 0 0 0 0 0 11 1 1 1 1

uhj(m1)

…. ….

k k

hi(m1)hj(m2)

hi(m2)

Decaying/Weakening the setofIDs

Randomly reset 1’s to 0. Same as EDBF [Kumar et al. Infocom’05]

Let (m)=i=1m B(hi(m))

Large (m) ! fresh mapping(m)/k is a rough measure of P{xm 2

A}

….0 1 0 0 0 0 0 0 11 1 1 1 1

hj(m) hk(m)h1(m) hi(m)

….0 1 0 0 0 0 0 0 11 1 1 1….0 1 0 0 0 0 0 0 11 1 1 1 1….0 1 0 0 0 0 0 0 11 1 1 1….0 1 0 0 0 0 0 01 0 1 1 1

Weakening State (Contd)

xn

2

uB

setofIDs small, time passes:Decay GeoRegion

xn

2

uB

xn

Either setofIDs large ORGeoRegion Large =>

Decay SetofIDs

Random Directional Walks

Both used to announce location information (“put”) and forward packets (“get”)

Dissemination/Proactive Phase: (put)

When a node receives a location announcement, it creates a ID-to-

location mapping aggregates this

mapping with previously created mappings if possible

A

B

C

Confidence0.71

Confidence0.84

Confidence1.0

Confidence0.71

Confidence0.84

Confidence0.71

Confidence0.84

Confidence0.71

Confidence0.71

Forwarding Packets(get)

Forwarding decision: similar to longest-prefix-match. “strongest semantics match” to decide how to bias the random walk.

Confidence0.71

Confidence0.84

Confidence1.0

1.0

Confidence0.71

Confidence0.84

Confidence1.0

1.0

Confidence0.71

Confidence0.84

Confidence1.0

S

D

A

B

C

E

WSR involves unstructured, flat, but scalable routing ; no flooding !

Simulation Objectives How does WSR perform?

Large-scale High Mobility

Comparisons: DSR: works well for small scale, high mobility GLS+GPSR: works well for large scale, low mobility

Short answer: 98%+ packet delivery despite large scale AND high mobility.

Tradeoffs: longer path lengths, (N3/2) state overhead

Simulation SetupBenchmarks

DSR and GLS-GPSRRandom waypoint mobility model with

vmin=5 m/s and vmax=10 m/sWSR is robust against dynamism (please see

the paper for details)Performance Metrics

Packet delivery ratioControl packet overheadNumber of states maintained Normalized path lengthEnd-to-end Delay

400 500 600 700 800 900 1000-0.2

0

0.2

0.4

0.6

0.8

1

1.2

Number of Nodes

Pac

ket

Dev

iver

y R

atio

WSR

GLS-GPSRDSR

Packet Delivery Ratio

GLS breaks down due to overheads

DSR only delivers a small fraction of

packets

WSR achieves high delivery ratio

Control Packet Overhead

400 500 600 700 800 900 10000

1000

2000

3000

4000

5000

6000

7000

Number of Nodes

Tot

al O

verh

ead

per

Sec

ond

(Num

ber

of P

acke

ts)

WSR

GLS-GPSR

Maintaining structure requires

superlinearly increasing

overhead in GLS

400 500 600 700 800 900 10000

0.5

1

1.5

2

2.5

3

3.5

4x 10

4

Number of Nodes

Tot

al N

umbe

r of

Map

ping

s/D

atab

ase

Ent

ries

Mai

ntai

ned WSR

GLS-GPSR

Number of States Maintained

The total state stored in the network increases as (N3/2) instead of

(NlogN)

Per-Node State Dynamics

0 100 200 300 400 500 600 700 800 900 10000

5

10

15

20

25

30

35

40

45

time (seconds)

Num

ber

of S

tate

s M

aint

aine

d

State generation rate matches state removal rate.

Distribution of Per-Node State in the Network

The states are well distributed with a C.o.V 0.101

20 25 30 35 40 45 50 550

20

40

60

80

100

120

Number of States

Num

ber

of O

ccur

renc

es

Normalized Path Length

400 500 600 700 800 900 10001

1.5

2

2.5

3

3.5

4

4.5

Number of Nodes

Nor

mal

ized

Pat

h Le

ngth

WSR

GLS-GPSRDSR

GLS sends packets only to destinations that are successfully located

Packets take longer paths with WSR

400 500 600 700 800 900 10000

10

20

30

40

50

60

70

Number of Nodes

End

to

End

Del

ay (

s)WSR

GLS-GPSRDSR

But, E2E Delay is Lower!

WSR uses random walks for discovery &

dissemination => end-to-end delay is smaller

Discussion/Future Work

Weak State Routing also relates toDTN routingUnstructured rare object recall in P2P

networksDistributed Hashing

Future work: Such extensions (especially DTNs)Theoretical analysis

Conclusion

Weak state is soft, updated locally, probabilistic and captures uncertainty

Random directional walks both for location advertisement and data forwarding.

WSR provides scalable routing in large, dynamic MANETs

WSR achieves high delivery ratio with scalable overhead at the cost of increased path length

Thank you

Questions?