XS-WSNetExtreme Scale Wireless Network Simulator

Asim Ali

Laboratoire de Recherche en InformatiqueBt 490 Universit Paris-Sud 1191405 Orsay Cedex France

ali.asim@lri.fr

May 15, 2009

Outline

Introduction WSNs.

Issues in WSNs.

Simulators.

WorldSens Environment.

XS-WSNet (Design, Implementation, Key Features).

Application Epidemic Protocol.

Experimental Results

Conclusions.

Questions/Comments.

Asim Ali (LRI) XS-WSNet May 15, 2009 2 / 21

Introduction to Wireless Sensor Networks

Smart devices with multiple sensors connected through wireless linksand internet.

Diverse applications like defense, security, traffic control, videosurveillance, environment monitoring, etc.

Overlaps with technologies from different fields e,g. sensing,computing, communication.

Sensors, processors, communication devices getting smaller andsmarter.

Reduction in size and price made possible large size networks.

Asim Ali (LRI) XS-WSNet May 15, 2009 3 / 21

Issues in WSNs

Limited storage, computing and communication capacity.

Unpredictable medium.

Unreliable wireless links.

Sensor nodes prone to failures.

Testing protocols and applications difficult.

Asim Ali (LRI) XS-WSNet May 15, 2009 4 / 21

Simulators (low level)

Also called emulators or platform simulators e.g. WSim, TOSSIM,ATEMU, etc.

Provides more details at hardware level and node internals.

Run the actual application code.

Very limited network scalability.

Asim Ali (LRI) XS-WSNet May 15, 2009 5 / 21

Simulators (high level)

High level or network simulators e.g. ns-2, GloMoSim, GTSNets,OMNet++, etc.

Provides more details at network and protocol levels.

Differs in design choices, implementation languages, underlyingoperating systems.

Limited scalability due to limited resources.

Only a few use parallelism e.g. GloMoSim.

Asim Ali (LRI) XS-WSNet May 15, 2009 6 / 21

WorldSens Environment

Two simulators WSNet(network simulator), WSim(sensor nodesimulator).

WSNet features: event driven, modular design, extensible.

Various models for different layers like application, MAC, radio,antenna, energy, etc.

Choice to configure the simulation environment.

Simulation configuration: in the form of XML file.

Asim Ali (LRI) XS-WSNet May 15, 2009 7 / 21

Example WSNet Simulation Configuration

Here is the small portion of an xml configuration file.

<!-- == Worldsens ===================================================== -->

<simulation nodes="10000" x="200" y="200" z="0"/>

<!-- == Entities ====================================================== -->

<!-- == PROPAGATION, INTERFERENCES and MODULATION ===================== -->

<entity name="range" library="propagation_range" > <init range="5"/></entity>

<entity name="interf" library="interferences_orthogonal"></entity>

<entity name="none" library="modulation_none"></entity>

<!-- == Environment ===================================================== --><environment><propagation entity="range" range="5"/><interferences entity="interf"/><modulation entity="none"/></environment>

</worldsens>

Figure: Small portion of an xml configuration file.

Asim Ali (LRI) XS-WSNet May 15, 2009 8 / 21

XS-WSNet Design Features

Basic idea to use parallelism using multiple machines.

Node/network paritioning scheme: divide nodes on number ofmachines with full area on each machine.

Communication module: developed and embedded with WSNet.

Changes in WSNet: made necessary changes in WSNet to work indistributed environment.

Asim Ali (LRI) XS-WSNet May 15, 2009 9 / 21

XS-WSNet Architecture

Figure: The XS-WSNet simulator on each machine

Asim Ali (LRI) XS-WSNet May 15, 2009 10 / 21

Application: Epidemic Protocol

Nodes distributed randomly in simulation area.

One source node: emits a message.

Sensor nodes receive and forwards the message with certainprobability. Do not forwrad the same message twice.

Communication mode: local broadcast.

Comparison criterias for single and multiple machine simulations:Number of receiving nodes.Average path length from source to sensor node.

Asim Ali (LRI) XS-WSNet May 15, 2009 11 / 21

Comparison: Number of Receiving Nodes

9600

9800

10000

10200

10400

0 2 4 6 8 10 12 14 16 18 20

no o

f rec

eive

nod

es

number of tests

wsnet comparison of total number of receive nodes for approximately 10000 nodes

100 nodes per machine200 nodes per machine500 nodes per machine

1000 nodes per machine10000 nodes per machine

Figure: Number of receiving nodes when simulated on different number ofmachines.

Asim Ali (LRI) XS-WSNet May 15, 2009 12 / 21

Comparison of Average Number of Hops

0

10

20

30

40

50

60

0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000

no o

f hop

s

number of nodes

wsnet comparison of avg number of hops for approximately 10000 nodes

100 nodes per machine200 nodes per machine500 nodes per machine

1000 nodes per machine10000 nodes per machine

Figure: Avgerage number of hops from source node to sensor node.

Asim Ali (LRI) XS-WSNet May 15, 2009 13 / 21

Exapmle Asynchronicity in XS-WSNet

Figure: All nodes on samemachine.

Figure: Node b1 on differentmachine.

Asim Ali (LRI) XS-WSNet May 15, 2009 14 / 21

Network Scalability

Network size: number of nodes.

Simulation time: difference of message receiving time of last nodeand emitting time of source node.

Node density: local node density, global node density.

Scale up options: increasing nodes on each machine or increasingnumber of machines.

Asim Ali (LRI) XS-WSNet May 15, 2009 15 / 21

Time compariosn: single vs five machines

0

200

400

600

800

1000

1200

0 20 40 60 80 100 120 140 160 180 200

time

in s

econ

ds

number of nodes x 1000

number of nodes vs time

single machine, area=300x300, range=55 machines, area=300x300, range=5

Figure: Simulation times up to 200,000 nodes.

Asim Ali (LRI) XS-WSNet May 15, 2009 16 / 21

Scalability with increasing node density

0

2000

4000

6000

8000

10000

12000

14000

16000

18000

100 200 300 400 500 600 700

time

in s

econ

ds

number of nodes x 1000

number of nodes vs time

single machine, area=300x300, range=55 machines, area=300x300, range=5

10 machines, area=300x300, range=5

Figure: Simulation times up to 700,000 nodes.

Asim Ali (LRI) XS-WSNet May 15, 2009 17 / 21

Scalability with increasing node density

2000

4000

6000

8000

10000

12000

14000

16000

18000

20000

22000

700 750 800 850 900 950 1000

time

in s

econ

ds

number of nodes x 1000

number of nodes vs time

single machine, area=1000x1000, range=55 machines, area=1000x1000, range=5

Figure: Simulation times from 700,000 to one million nodes.

Asim Ali (LRI) XS-WSNet May 15, 2009 18 / 21

Scalability with constant node density

0

10000

20000

30000

40000

50000

60000

70000

80000

90000

1 2 3 4 5 6 7 8 9 10

time

in s

econ

ds

number of nodes in millions

wsnet scale up times upto 10 million nodes with constant node density

number of machines = 1number of machines = nodes in millions * 5

Figure: Simulation times from one million up to 10 million nodes.

Asim Ali (LRI) XS-WSNet May 15, 2009 19 / 21

Conclusions

Proposed, implemented and tested a distributed version of WSNetsimulator with scalability as prime objective.

XS-WSNet provides scale up and speed up even with small numberof machines.

Scalability is linear with number of participating machines.

Provides asynchronous environment for simulation.

Asim Ali (LRI) XS-WSNet May 15, 2009 20 / 21

Any Questions/Comments

Thanks.Any questions please.

Asim Ali (LRI) XS-WSNet May 15, 2009 21 / 21