1. Network Modeling and Simulation with Network Simulator 2 (ns2)
Department of Computer Science and Engineering
University of Bologna. Italy
Marco Di Felice

2. Outline NS2: Use Cases and Features NS2: Architecture and Design
NS2: Building simple network models
NS2: Demo & Analysis

3. Ns2: An Overview
NS2: A (discrete event) network simulator tool.
Generally speaking, a network simulator is a dedicated software that allows to:
Model the behaviour of network protocols/applications (e.g. TCP protocol).
Reproduce the behaviour of a computer network as a whole (e.g. a wireless LAN).
Quantify the network performance, through well-defined network metrics (e.g. system throughput).

4. Ns2: WHAT Case 1: Network planning
Q. When may I use a network simulator?
Case 1: Network planning
Need to build a network infrastructure, with coverage and Quality of Service (QoS) issues.
Q1. Where to place the
network gateways?
Q2. How many gateways
do I really need?

5. An example: Early Collision Detection Systems In VANETs
Ns2: WHAT
Q. When may I use a network simulator?
Case 2: Network Evaluation
An example: Early Collision Detection Systems In VANETs
Need to evaluate the performance of a (possibly large-scale) network infrastructure or network application.
Use an existing testbed  BUT …
Q1. What is the cost of setting up the experiments?
Q2. Are the experiments easily reproducible?

6. An example: Proposing a new TCP variant for wireless LANs.
Ns2: WHAT
Q. When may I use a network simulator?
Case 3: Research on Networking Systems
Need to evaluate the performance of a new network protocol, architecture, or application.
An example: Proposing a new TCP variant for wireless LANs.
1. A real deployment might not be feasible, or might be too costly.
2. Analytical models might be too complex to model all the components of the scenario under study.

7. Ns2: WHAT
Network simulators might be used to model several kinds of networking systems (wired, wireless, optical, etc).
In practice, simulation constitutes the main evaluation technique for wireless systems.
Possibility to build reproducible experiments (hard to guarantee with wireless testbeds).
Possibility to reproduce wireless propagation phenomena in an accurate way through probabilistic models (e.g. fading)
Possibility to model large-scale wireless networks composed by several interacting nodes.

8. Ns2: WHAT
Simulation is a meaningful evaluation approach only when it produces “trustable” results.
Validation is needed to certify that the simulation models reproduce correctly the characteristics and dynamics of the system under study.
HOW to VALIDATE a NETWORK MODEL?
Compare Simulation vs Analytical Model
Compare Simulation vs Real Measurements
SIM.
THROUGPUT
EXPERIMENTS
ANALYTICAL

9. Ns2: WHAT Discrete Event simulator (details later …)
NS2: A network simulation tool
Discrete Event simulator (details later …)
NS2 allows to model and evaluate several IP networking systems (LAN/WAN). It includes:
Network protocols model (e.g. MAC, routing, …)
Network applications model (e.g. CBR, FTP, …)
Queue management algorithms (e.g. FIFO, RED, …)
Network link models (e.g. lossy link) …

10. Ns2: WHERE http://www.isi.edu/nsnam/ns
download ns-allinone
includes several tools (ns2, nam, awk, otcl, …)
mailing list:
documentation:
Reference manual and Tutorials on the website
Other tutorials on the Web

11. Ns2: WHEN The project started from REAL project in 1989
ns-1 by Floyd and McCanne at Lawrance Berkeley National Laboratory (LBNL).
ns-2 by Steve McCanne, within the VINT project involving a consortium of US universities (LBL, PARC,USC, ...)
currently maintained at USC/ISI (University of Southern California), but several forks available.
NS3 relased in 2008 (now NS3.15)
Deployed by a team lead by Tom Henderson and Sally Floyd (University of Washington)
A completely new tool, not a mere extension of Ns2 !

12. Ns2: WHY Discrete Event simulator (details later …) OMNET++ OPNET JiST
NS2: A network simulation tool
Discrete Event simulator (details later …)
OTHER SIMULATION TOOLS
OMNET++
OPNET
JiST
NS3
GLOMOSIM …
Q. WHY should I use NS2 for my research?

13. Ns2: WHY NS2 includes a vast model library of network components.
Link Models:
Wired Links,
Wireless Links,
Satellite Links, …
Application Layer
FTP, Telnet, HTTP, Multimedia, Exponential traffic, …
Transport Layer
UDP, TCP (Reno, NewReno, Vegas, SACK), …
Network Layer
Wired routing (RIP), Ad Hoc Routing (AODV, OLSR, DSR), …
MAC Layer
Ethernet, (WIFI), , Bluetooth, (WIMAX), …

14. Ns2: WHY Collaborative deployment environment
NS2 is distributed with GNU General Public Licence (GPL)
Collaborative deployment environment
possibility to freely modify the existing NS2 code based on each user’s needs.
possibility to share network models for research/education purposes (e.g. a new implementation of TCP).
possibility to compare his/her own model with models implemented by other research teams.
Multi-platform support
GNU/Linux, MAC OSX, Solaris, Windows, …

15. Ns2: WHY
NS2 is a popular tool, widely adopted by researchers working on the field of computer networks.
Number of Users: 10K
Institutes: 1K
% Papers: 44.4%
MOBIHOC Conference: Statistics on tools used to produce a simulation study within the papers submitted at the ACM MOBIHOC conference ( ).

16. Ns2: WHY ? NS2 Q. WHY NOT to use NS2 for my research?
Performance issues
Monolithic (basic) scheduler, scalability constitutes a big issue.
Architectural issues
Not really a modular architecture, difficult to share the code of network models.
Evaluation issues
No support for the trace analysis.
CPU time ?
#nodes
NS2
TRACE

17. NS2: HOW Two programming languages: C++ and OTCL.
OTCL for simulation setup and execution
Quickly define the simulation environment
C++ for model deployment
Implement the behaviour of a network component

18. that should be invoked once the event is fired.
NS2: HOW
The core of the NS2 simulator is the Scheduler
Discrete-event scheduler.
Basic implementation, few optimization.
Event in NS2 ID
TYPE
TIME
HANDLER
Packet sent
NS2 Object
+ function name
that should be invoked once the event is fired.
Packet
Events
Timer
Events
Packet received
Packet dropped
Simulation time of the event

19. NS2: HOW
The Scheduler manages the simulation event list.
The elements are the events of the simulation.
The list is ordered on the basis of the time field. E1 E2 E3 E4 E5 E6
SIMULATION
TIME: t
SIMULATION
TIME: E1.time
At each simulation step:
The head element of the list is removed
The simulation time is set to E1.time
The event handler (E1.handler) is executed. 1
TYPE
TIME
HANDLER

20. NS2: HOW
The Scheduler manages the simulation event list.
The elements are the events of the simulation.
The list is ordered on the basis of the time field. E2 E3 E4 E5 E6 E7
SIMULATION
TIME: E1.time
SIMULATION
TIME: t
At each simulation step:
E1.handler is executed, and it might generate new events (e.g. E7), that are inserted into the event list (at a position denoted by E7.time)
E1.HANDLER

21. NS2: HOW Let’s make an example on a network scenario …
NODE A
NODE B
APPLICATION
APPLICATION
MAC
MAC
ETHERNET
LINK
SIMULATION
TIME: 0
EVENT LIST
At t=0, the Application module of node A is invoked

22. NS2: HOW Let’s make an example on a network scenario … E1
NODE A
NODE B
APPLICATION
APPLICATION
MAC
MAC
ETHERNET
LINK
SIMULATION
TIME: 0 E1
EVENT LIST 1
Send 4
A.APPLICATION
A timer event is scheduled at time 4 by node A

23. NS2: HOW Let’s make an example on a network scenario … E2 E3 2 Recv
NODE A
NODE B
APPLICATION
APPLICATION
MAC
MAC
ETHERNET
LINK
SIMULATION
TIME: 4 E2 E3
EVENT LIST 2
Recv
4.5
A.MAC 3
Send 8
A.APPLICATION

24. NS2: HOW Let’s make an example on a network scenario … E4 E3 4 Recv
NODE A
NODE B
APPLICATION
APPLICATION
MAC
MAC
ETHERNET
LINK
SIMULATION
TIME: 4.5 E4 E3
EVENT LIST 4
Recv
5.0
B.MAC 3
Send 8
A.APPLICATION

25. NS2: HOW Let’s make an example on a network scenario … E5 E3 5 Recv
NODE A
NODE B
APPLICATION
APPLICATION
MAC
MAC
ETHERNET
LINK
SIMULATION
TIME: 5 E5 E3
EVENT LIST 5
Recv
5.2
B.APPLICATION 3
Send 8
A.APPLICATION

26. NS2: HOW Let’s make an example on a network scenario … E3
NODE A
NODE B
APPLICATION
APPLICATION
MAC
MAC
ETHERNET
LINK
SIMULATION
TIME: 5.2 E3
EVENT LIST 3
Send 8
A.APPLICATION
The message is processed by Node B at time 5.2

27. Ns2: HOW Two ways of interactions: Modify/Create a new network model
Network models: network protocols, applications, queue policies, network architecture models, etc.
Coding in C++
Recompile at the end.
Configure/Run a network simulation
Coding in OTCL
Executed by an interpreter, no need to recompile.
NOT EASY
QUITE EASY

28. Ns2: HOW Running an OTCL script:
ns script-file.tcl [parameters]
Initialize the scheduler
Define the simulation parameters (e.g. start time)
Build the network topology
Generate the traffic load
Define the protocol stack used by each node
OTCL  scripting language, OO-extension of TCL 

29. Ns2: OTCL inside Assign a value to a variable
set x 0
Keyword $ returns the value of a variable
set y $x
Selection Statements if (if < expr > ... else ...)
if {$x == $y } { puts “Hello world” }
Iterative Statements
for {set i 0; $i < $x ; incr i}{puts “Hello world” }
Function Declaration
proc name_FUNCTION {par1, ...parn} {... return $x}
OTCL Overview

30. Ns2: HOW Running an OTCL script:
ns script-file.tcl [parameters]
Initialize the scheduler
Define the simulation parameters (e.g. start time)
Build the network topology
Generate the traffic load
Define the protocol stack used by each node
OTCL  scripting language, OO-extension of TCL 

31. Ns2: Initialize the Scheduler
Creating the Event Scheduler
set ns [new Simulator]
Starting the simulation
$ns run
Initializing the random number generator
$ns-random SEED
Scheduling the events
$ns at <time> <event>
Stopping the simulation at time 300
$ns at 300 "finish“
SEED=0  current timestamp
All the random variable used by the current simulation are initialized with this SEED.

32. Ns2: HOW Running an OTCL script:
ns script-file.tcl [parameters]
Initialize the scheduler
Define the simulation parameters (e.g. start time)
Build the network topology
Generate the traffic load
Define the protocol stack used by each node
OTCL  scripting language, OO-extension of TCL 

33. Ns2: Building the network (WIRED)
CASE 1. Modeling a wired network.
Define the nodes of the network
set n0 [$ns node]
set n1 [$ns node]
Define the Links among nodes
#Nodes connected with an Ethernet cable, 10 Mb/s
$ns duplex-link $n0 $n1 10Mb 100ms DropTail
Specifies bandwidth, delay, and queue policy:
DropTail, RED, CBQ, FQ, SFQ, DRR

34. Ns2: Building the network (WIRED)
CASE 1. Modeling a wired network.
Define the error model on wired links
set loss_module [new ErrorModel]
$loss_module set rate_ 0.1
$loss_module ranvar [new RandomVariable/Uniform]
$loss_module drop-target [new Agent/Null]
$ns lossmodel $loss_module $n0 $n1
Lossy link between node 0 and node 1, with
error rate equal to 0.1. Packets with errors are sent
to Agent/Null, i.e. they are discarded.

35. Ns2: Building the network (WIRED)
CASE 1. Modeling a wireless network.
Define the nodes of the network
set n0 [$ns node]
set n1 [$ns node]
Define the position
set topograpy [new Topography]
$topography load_flatgrid
$n0 set X_ 300
$n0 set Y_ 400
$n0 set Z_ 0
Set simulation area
to 400mx400m
Set node 0 at position <300,400,0>

36. Ns2: Building the network (WIRED)
CASE 1. Modeling a wireless network.
Define the mobility of wireless nodes
NS_OBJ at TIME “NODE setdest X_COOR Y_COOR SPEED”
$ns at 10.5 “$node(0) setdest ”
At time 10.5, node 0 will move toward
position (100,100) with speed equal to
5 m/s (constant speed)
Utilize the General Object Director (GOD)
set $god [new God]
Object that stores global information
about the state of the environment (e.g.
the matrix of connectivity among nodes)

37. Ns2: Building the network (WIRED)
CASE 1. Modeling a wireless network.
The mobility traces of wireless nodes can be pre-generated by using the setdest tool (random waypoint model)
./setdest [-n num_of_nodes] [-p pausetime] [-maxspeed]
[-t simtime] [-x][-y] > [fileOutput]
In the TCL script: source “fileOutput”
Any mobility simulator can be used for trace generation.
MOBILITY
SIMULATOR
MOB.
TRACE
OTCL
SCRIPT
NS2
e.g. SUMO
SOURCE

38. Ns2: HOW Running an OTCL script:
ns script-file.tcl [parameters]
Initialize the scheduler
Define the simulation parameters (e.g. start time)
Build the network topology
Generate the traffic load
Define the protocol stack used by each node
OTCL  scripting language, OO-extension of TCL 

39. Ns2: Creating connections (UDP/TCP)
Define the end-points of the communication
TCP Connections:
set src [new Agent/TCP]
set dst [new Agent/TCPSink]
UDP Connections:
set src [new Agent/UDP]
set dst [new Agent/Null]
Connect sender and receiver
$ns attach-agent $n0 $src
$ns attach-agent $n1 $dst
$ns connect $src $dst
Several TCP variants:
TCP Tahoe
TCP Reno
TCP NewReno
TCP Vegas
TCP SACK …

40. Ns2: Attaching Applications
Define the application and attach it to the sender
FTP Agent
set ftp [new Application/FTP]
$ftp attach-agent $src
$ns at TIME “$ftp start”
CBR Agent
set cbr [new Application/Traffic/CBR]
$cbr attach-agent $src
$ns at TIME “$cbr start”
Exponential Traffic Generator
set exp [new Application/Traffic/Exponential]

41. Ns2: HOW Running an OTCL script:
ns script-file.tcl [parameters]
Initialize the scheduler
Define the simulation parameters (e.g. start time)
Build the network topology
Generate the traffic load
Define the protocol stack used by each node
OTCL  scripting language, OO-extension of TCL 

42. Ns2: HOW
A wireless environment can be modeled by configuring the protocol stack of each node.
$ns_ node-config –phyType $val(netif)
-propType $val(prop)
-antType $val(type)
-llType $val(ll)
-macType $val(mac)
-ifqType $val(ifq)
-ifqLen $val(ifqlen)
-adhocRouting $val(rp)
-topoInstance $topo
-channel $chan_
NET LAYER
QUEUE
LL LAYER
MAC LAYER
ANTENNA
PROPAGATION
PHY LAYER

43. Ns2: HOW
A wireless environment can be modeled by configuring the protocol stack of each node.
$ns_ node-config –phyType $val(netif)
-propType $val(prop)
-antType $val(type)
-llType $val(ll)
-macType $val(mac)
-ifqType $val(ifq)
-ifqLen $val(ifqlen)
-adhocRouting $val(rp)
-topoInstance $topo
-channel $chan_
NET LAYER
QUEUE
LL LAYER
MAC LAYER
ANTENNA
PROPAGATION
PHY LAYER

44. Ns2: HOW Configuring the PHY Layer set val(netif) Phy/WirelessPhy[Ext]
Some parameters to be tuned:
Phy/WirelessPhy set Pt e-01
Phy/WirelessPhy set RXThresh e-08
Phy/WirelessPhy set CSThresh e-09
Functionalities offered by the PHY Layers
Signal capture
Modulation & Bit-rate setting
Modeling of collision/transmission errors …

45. Ns2: HOW
A wireless environment can be modeled by configuring the protocol stack of each node.
$ns_ node-config –phyType $val(netif)
-propType $val(prop)
-antType $val(type)
-llType $val(ll)
-macType $val(mac)
-ifqType $val(ifq)
-ifqLen $val(ifqlen)
-adhocRouting $val(rp)
-topoInstance $topo
-channel $chan_
NET LAYER
QUEUE
LL LAYER
MAC LAYER
ANTENNA
PROPAGATION
PHY LAYER

46. Ns2: HOW Configuring the Propagation model
set val(prop) Propagation/TwoRayGround
set val(prop) Propagation/FreeSpace
FREE SPACE
TWORAY
RECEIVER
SENDER
Configuring the Antenna model
set val(antType) Antenna/OmniAntenna
set val(antType) Antenna/Directional
OMNIDIRECTIONAL
DIRECTIONAL

47. Ns2: HOW
A wireless environment can be modeled by configuring the protocol stack of each node.
$ns_ node-config –phyType $val(netif)
-propType $val(prop)
-antType $val(type)
-llType $val(ll)
-macType $val(mac)
-ifqType $val(ifq)
-ifqLen $val(ifqlen)
-adhocRouting $val(rp)
-topoInstance $topo
-channel $chan
NET LAYER
QUEUE
LL LAYER
MAC LAYER
ANTENNA
PROPAGATION
PHY LAYER

48. Ns2: HOW Configuring the LL layer Configuring the MAC model
set val(ll) LL
Include ARP protocol
Configuring the MAC model
set val(mac) Mac/802_11
Select a MAC protocol:
(Wifi)
(Sensors)
CSMA/CA …
Configuring the Queue Layer
set val(ifq) Queue/DropTail/PrimaryQueue
set val(ifqlen) 50
Define the queue policy:
PrimaryQueue
RED Queue …
Set the queue length

49. Ns2: HOW
A wireless environment can be modeled by configuring the protocol stack of each node.
$ns_ node-config –phyType $val(netif)
-propType $val(prop)
-antType $val(type)
-llType $val(ll)
-macType $val(mac)
-ifqType $val(ifq)
-ifqLen $val(ifqlen)
-adhocRouting $val(rp)
-topoInstance $topo
-channel $chan
NET LAYER
QUEUE
LL LAYER
MAC LAYER
ANTENNA
PROPAGATION
PHY LAYER

50. Ns2: HOW Configuring the routing protocol set val(adhocrouting) AODV
Select a routing protocol for multi-hop networks:
AODV, DSDV, DSR, TORA, ….
SOURCE
ROUTING PATH
DESTINATION

51. Ns2: HOW Two ways of interactions: Modify/Create a new network model
Network models: network protocols, applications, queue policies, network architecture models, etc.
Coding in C++
Recompile at the end.
Configure/Run a network simulation
Coding in OTCL
Executed by an interpreter, no need to recompile.
NOT EASY
QUITE EASY

52. Ns2: HOW In C++, each model extends the class NSObject.
Each NSObject has a correspective in OTCL.

53. Ns2: HOW When creating a new model in C++: Extend the NSObject class
Create the corresponding OTCL class
Implement these methods
recv(Packet* p, Handler* h)  Callback once a packet is received from the upper layer.
command(int argc, const char*const* argv)  Binding between C++ and OTCL for the parameter passing from the TCL script.

54. Ns2: Simulation Output (TRACE)
The output of the simulation is a trace file, containing the description of the events occurred during the simulation.
s MAC RTS 44 [253e 1 0 0]
r MAC RTS 44 [253e 1 0 0]
s MAC CTS 38 [ ]
r MAC CTS 38 [ ]
s MAC cbr 1112 [13a ]
r MAC cbr 1112 [13a ]
Simulation
Time
Node
Packet
size
Traffic
type
MAC Header
Packet ID
Event Type

55. Ns2: Simulation Output (TRACE)
Depending on the length of the simulation, the trace file might occupy lots of bytes on the disk.
$ns_ node-config –agentTrace ON/OFF
-routerTrace ON/OFF
-macTrace ON/OFF
-mobilityTrace ON/OFF
Configure the granularity of the tracing process …
s AGT cbr 1112 [13a ]
s MAC cbr 1112 [13a ]
r MAC cbr 1112 [13a ]
r AGT cbr 1112 [13a ]

56. Ns2: Simulation Output (NAM)
The output of the simulation can be visualized by using the Network Animator (NAM) tool.

57. Ns2: Analysis of Simulation Results
NS2 does NOT provide any specific support for the data analysis/validation and for the computation of performance metrics (e.g. throughput, delay).
Run multiple simulations with different seeds
Remove the transient phase from the trace file
Extract the performance metrics from the trace file
Compute the average and confidence intervals
Plot the results
External data processing tools must be used.
GNUPLOT
AWK, PERL, …

58. Ns2: Analysis of Simulation Results
Example: Computing the system throughput in AWK.
BEGIN {
recvByte=sim_time=transient=0.0 }
($1==‘r’) && ($4==‘AGT’) && ($2>transient) {
recvByte+=$8
sim_time=$2
END {
print recvByte/(sime_time-transient)