the price of mobility in adhoc...

The Price of Mobility in AdHoc Networks

Dr Kim BlackmoreDepartment of Engineering

Australian National University

The Price of Mobility in AdHoc Networks AusCTW’05 1

What is a MANET?

➤ Ad-Hoc: no central administration.All nodes can act as routers and hosts.Nodes independently and spontaneously form a network.

➤ MANET: All nodes are mobile and use wireless interfaces to send packetdata.

➤ Dynamic network topology


Routing

➤ Wired network well establishedGuarantee optimally efficient routes

➤ Cellular network — rely on fixed infrastructureLocation managementHandoff managementOptimality vs overhead

➤ Adopting wired protocols to MANETs doesn’t workRepeatedly recalculating optimal routes


Conventional Protocols

➤ Link State Routing

➤ Distance Vector Routing

➤ Source Routing


Classification of MANET routing protocols

➤ Multi-channel or Single-channel

➤ Static or Adaptive

➤ Centralized or Distributed

➤ Uniform or Non-uniform

➤ Topology-based or Destination-based

➤ Proactive or Reactive

➤ Cost function: hop count; bandwidth; energy; degree of association stability


A list of MANET Routing Protocols

➤ CGSR - Clusterhead Gateway Switched Routing

➤ DSR - Dynamic Source Routing

➤ DSDV - Destination Sequenced Distance Vector

➤ WRP - Wireless Routing Protocol

➤ AODV - AdHoc On-Demand Distance Vector

➤ TORA - Temporally Ordered Routing Algorithm

➤ OLSR - Optimized Link State Routing

➤ ZRP - Zone Routing Protocol

➤ CEDAR - Core Extraction Distributed Ad hoc Routing


Evaluating Routing Algorithms

➤ Simulate

➤ NS-2

➤ GloMoSim

➤ MAISIE

➤ OPNET Modeler

➤ Study how performance changes as network behaviour gets morecomplicated.

➤ Performance Metrics

➤ Topological Change (also Traffic Change)

➤ Mobility Metrics


Performance Metrics

➤ Packet Delivery Ratio = Throughput

➤ Protocol Overhead

➤ Average End-to-End Delay


Topological change for simulations

Simulations test routing protocol under a particular dynamic topology.

Topological change is determined by

1. Nodes moving

2. Multipath Fading of channels between nodes

3. Time varying noise and interference

Common to consider only node movement, and assume free-spacepropagation between nodes.

Links are assumed to be on or off, and bidirectional.


Performance Comparison of Common Algorithms

0 10 20 30 40 50 60Maximum Speed (m/sec)

50

60

70

80

90

100

Thr

ough

put (

%)

DSDVAODVDSR


0

2000

4000

6000

8000

10000

Rou

ting

Ove

rhea

d (p

acke

ts)

DSDVAODVDSR

Clearly DSR is the winner. . .

Figures from ”IMPORTANT: A framework to systematically analyze the Impact of Mobility on Performance Of

RouTing in Ad-hoc NeTworks”, F. Bai, N. Sadagopan, A. Helmy*, InfoCom2003. Used with permission.


Mobility Models

➤ Random Walk Mobility Model

➤ Random AdHoc Mobility Model

➤ Random Waypoint Mobility Model

➤ Gauss-Markov Mobility Models

➤ City Section Mobility Models

➤ Freeway Mobility Model

➤ Reference Point Group Model

➤ Swarm mobility models


The Problem with mobility models


50

60

70

80

90

100

Thr

ough

put (

%)

DSDVAODVDSR


50

60

70

80

90

100

Thr

ough

put (

%)

DSDVAODVDSR

Which protocol has the highest throughput?




The Problem with mobility models(2)


0

2000

4000

6000

8000

10000

Rou

ting

Ove

rhea

d (p

acke

ts)

DSDVAODVDSR


0

25000

50000

75000

1e+05

1.25e+05

1.5e+05

Rou

ting

Ove

rhea

d (p

acke

ts)

DSDVAODVDSR

Which protocol has the lowest overhead?




Mobility Metrics

➤ Node Degree

➤ Relative Speed

➤ Spatial Dependence

➤ Temporal Dependence

➤ Link Duration

➤ Link Availability

➤ Link Available time

➤ Path Duration. . .


What makes a good mobility metric?

➤ Computable in a distributed environment without global networkknowledge.

➤ A good indicator of protocol performance

➤ Feasible to compute (in terms of node resources)

➤ Independent of any specific protocol

➤ Computable in real network implementations


Does link duration quantify topological change

4.5 5 5.5 6 6.5 7 7.5 8 8.5 910

15

20

25

30

35

40

45

50Packet Delivery Ratio Using Random Walk Mobility Model

Pac

ket D

eliv

ery

Rat

io (

%)

Average Link Duration (seconds)

AODVDSRDSDVTORA

2 3 4 5 6 7 8 9 10 1110

20

30

40

50

60

70

80Packet Delivery Ratio Using Random Waypoint Mobility Model

Pac

ket D

eliv

ery

Rat

io (

%)

Average Link Duration (seconds)

AODVDSRDSDVTORA

Which protocol has the highest packet delivery ratio?


More problems with simulations

Different simulators do not give the same results.

0

10

20

30

40

50

60

70

80

90

100

50 100 150 200 250 300

Suc

cess

rat

e [p

erce

nt]

Power range [meters]

OPNET ModelerNS-2

GloMoSim0

500

1000

1500

2000

2500

3000

0 5 10 15 20

Ove

rhea

d [p

acke

ts]

Max speed [meters/second]

OPNET ModelerNS-2

GloMoSim

Figures from: D. Cavin, Y. Sasson, A. Schiper, ”On the Accuracy of MANET Simulations”, Principles of Mobile

Computing 2002, Toulouse, France, 2002. See ACM. Used with permission.


Beyond simulation

Entropy imposes a minimum overhead.

This restricts the capacity of the network, and thus the maximum throughput.

Should give theoretical bounds on the potential performance of AdHocnetworks, to use for benchmarking simulation results.

See poster ”On entropy measures for dynamic network topologies: Limits toMANET”, Roy Timo, Kim Blackmore and Leif Hanlen.

See also ”Path availability in mobile ad hoc networks”, Sanlin Xu, KimBlackmore and Haley Jones.


the price of mobility in adhoc...

Documents