Uwicore, Ubiquitous Wireless Communications Research Laboratory

Universidad Miguel Hernández, www.uwicore.umh.es

ns-3 workshopRome – March 2nd, 2009

ns-3 scalability constraints in heterogeneous wireless

simulations: iTETRIS a case study

Ramon Bauza (rbauza@umh.es), Javier Gozalvez (j.gozalvez@umh.es) and

Miguel Sepulcre (msepulcre@umh.es)

1

ns-3

work

shop –

Rom

e M

arc

h 2

009

Contents

� iTETRIS project presentation

� WAVE/802.11p

� ns-3 scalability capabilities

� Impact of number of vehicles and vehicle density

� Reduction of physical layer accuracy

� Reduction of interference range and packet rate

� Conclusions

2

ns-3

work

shop –

Rom

e M

arc

h 2

009

iTETRIS presentation

� iTETRIS (an Integrated Wireless and Traffic Platform for Real-Time

Road Traffic Management Solutions)

� European project funded within the 7th Framework Programme

� Consortium of different research groups: THALES, CBT, City of Bologne,

DLR, Eurecom, Hitachi, Innovalia, Peek Traffic and UMH

� Main iTETRIS objectives

� Implement an open-source integrated wireless and traffic simulation

platform

� Estimate the impact of cooperative vehicular

communications on traffic management

� Test and optimize V2V and V2I

communications and networking protocols

� Test and optimize cooperative traffic

management policies

� Large scale trials (traffic data of the city of

Bologne)

3

ns-3

work

shop –

Rom

e M

arc

h 2

009

iTETRIS simulation platform

� Integration of two widely used open source platforms

� ns-3 (Network Simulator 3)

� SUMO (Simulation of Urban MObility)

� CTMC (Cooperative Traffic Management Centre)

� Decision on routing traffic

flows

� Inform vehicles using the

MxC functional block

� MxC (Message eXchange Communications)

� Also provides the CTMC with traffic condition estimates (derived from

V2X communication)

4

ns-3

work

shop –

Rom

e M

arc

h 2

009

iTETRIS wireless platform architecture

� The iTETRIS wireless platform will follow the European ITS communications architecture determined by COMeSafety.

� This architecture has many similarities with the CALM architecture developed under the ISO.

Applications

Management

Facilities

Transport & Network

Access Technologies

Road Safety Traffic EfficiencyValue-Added

Servicies

Application

SupportSession Support

Information

Support

ITS

Network

ITS Transport

Geo-

Routing

Other

Protocols

Ipv6 +

Mobility

TCP/UDP

NF-SAP

FA-SAP

MA-SAP

MF-SAP

MN-SAP

MI-SAP

IN-SAP

WAVE UMTS WiMAX DVB

5

ns-3

work

shop –

Rom

e M

arc

h 2

009

WAVE/802.11p

� Spectrum divided into three 10MHz sub-channels

� WAVE MAC/PHY for CCH and SCH

6

ns-3

work

shop –

Rom

e M

arc

h 2

009

ns-3 scalability capabilities

� ns-3 advantages over ns-2� Good scalability, modularity and multi-technology (from the beginning)

� ns-3 performance for large scale simulations� Capable to simulate large amount of nodes (20000 nodes or more)

� High execution times for large and dense scenarios

� Work needed to achieve feasible runtimes for iTETRIS� Paralellization techniques (planned for June 2009)

� Simplification of ns-3 communication modules

� Factors directly impacting on simulation performance� Total number of vehicles

� Traffic density

� Implementation accuracy

� Interference range

� Packet generation rate

7

ns-3

work

shop –

Rom

e M

arc

h 2

009

Impact of number of vehicles and vehicle density

40sSimulation time

6MbpsData rate

10 packets/second (default)Packet tx rate

700m (default) Interference range

LogDistancePropagationLossPropagation model

RandomWalk2dMobilityModelMobility model

70km/h (maximum)Nodes’ speed

ValueParameter

� Scenario under evaluation

� 802.11 one-hop periodic broadcast

� Influence of vehicle density on simulation performance

� Close related to number of neighbours (packets to be

processed)

� Simulations with different traffic densities

700m

Inte

rfere

nce

rang

e

8

ns-3

work

shop –

Rom

e M

arc

h 2

009

Impact of number of vehicles and vehicle density

� Impact of traffic density on execution time (40s 5000 nodes)

� Execution time linearly increases

with traffic density

0 500 1000 1500 20000

5

10

15

20

25

30

35

Execution tim

e [h]

Vehicle density [Veh/Km2]

0 0.5 1 1.5 2

x 104

0

20

40

60

80

100

120

Exe

cu

tio

n tim

e [h

]

Number of vehicles

D10 (High Density)

D100 (Low Density)

� Influence of traffic density and number of vehicles on simulation performance

� Unworkable execution times for 1h

simulation

� Improvements need to be introduced

into the simulator

170Low

397High

Execution time [days]Traffic density

1h 20000 nodes

9

ns-3

work

shop –

Rom

e M

arc

h 2

009

Reduction of physical layer accuracy

� ns-3 spends most of the time at the physical layer

� Packets processing and interference calculation

� Simplification of PHY models

� Removal of interference calculation

� Performance analysis

� Around 35% of reduction time

40s High Density (D100)

0 0.5 1 1.5 2

x 104

0

20

40

60

80

100

120

Exe

cu

tio

n tim

e [h

]

Number of vehicles

Default

Without Interference

10

ns-3

work

shop –

Rom

e M

arc

h 2

009

Reduction of interference range and packet rate

� Congestive scenarios

� Vehicles adapt transmission range and packet rate to reduce interference40s Default PHY model

100 200 300 400 500 600 7000

20

40

60

80

100

120

Exe

cu

tio

n tim

e [h

]

Interference range [m]

10000 nodes

20000 nodes

� Impact of interference range on simulation performance

� Execution times increases with interference range

� Impact of packet rate on simulation performance

� Simulations run 80% faster reducing transmission rate from 10 to 2 packets per second

� Realistic congestive scenario� High density (D10)

� Interference range of 100m

� 2 packets per second

� 1h simulation with 20000 nodes would take 30 days

� Execution time much more feasible for iTETRIS 100m

400m

700m

Interference range

149

245

397

Estimated execution

time [days]

1h Default PHY 20000 nodes High Density

11

ns-3

work

shop –

Rom

e M

arc

h 2

009

Conclusions

� iTETRIS presents demanding requirements in terms of simulation platformscalability

� Multi-technology platform

� Large-scale scenarios

� Long simulation times to obtain valid results

� ns-3 capable of simulating large scale scenarios and high traffic densities

� Default distribution takes considerable time

� Need for future enhancements for optimization

� More efficient scheduler

� Parallelization techniques

� Staged simulation techniques

� Grid-based decomposition

� Code performance improvement

� Further suggestions and proposals