Cairo University Computer Networks-1

Faculty of Engineering CMP405-A /CMPN405

Computer Department Fall 2021

Network topologies with

omnet++
1
2 Today’s goal
 Build a LAN of an arbitrary N nodes.
 Any 2 nodes can send messages to each other (directly or indirectly).
 Experiment with different network topologies

Star Mesh Ring

3 We’ll learn about

1. Gate vectors
2. Dynamic connections
3. Sampling from random distributions
4. Self messaging
4
Gate vectors
Building gates of different sizes
5
Gate vectors

• Can create single gates and gate

vectors.
• Gate vectors follow the square
brackets syntax “[]”
Gate vector
• The size of a gate vector can be
Gate
specified or be left empty
• If left empty, size is inferred from
the connections, or is determined
when instantiated as a submodule
• You can find out the size of a gate by
calling gateSize(<gate_name>)
6 If size is left empty

Size is inferred Size is specified when

instantiated

Node1 has gate size 2, nodes 2 & 3 have a

size of 1
7
Gate vectors cont.

• Gate vectors can

accommodate and expand to
grow their pre-determined
size
• This code runs properly
without any errors.
• All nodes have a gate size of 2
8
Gate vectors cont.

We can also specify the size by

subclassing.
10 Gate vectors cont.

 Omnetpp by default doesn’t allow for a gate to be unconnected.

 An error will be produced when attempting to run with a gate that is unconnected.
 We can work around this by using the @loose property
11
Dynamic connections
12 Dealing with dynamic networks

 We can use for loops and if conditions inside connections to deal with arbitrarily
large modules and add some logic to our topology.
13 Dynamic networks cont.

N=4 (default) N=8

14 Nested loops
15
Simulating traffic
16 Simulating traffic

 We want to simulate traffic in our network.

 We can model a packet’s arrival at the sender node with a Poisson process, similar
to how we model queues.
 Then sampling from an Exponential distribution with a parameter lambda, we can
get the time interval for the next packet arrival.
 Upon arrival, we can then sample from a Uniform distribution a value to
determine which node our packet will be sent to.
17 Random sampling
 Omnetpp offers utility functions to sample from many different distributions.
 The two we’ll use the most are uniform and exponential as follows:
 uniform(double a, double b)

 exponential(double mean)
 Remember the exponential distribution’s mean = 1/lambda
 Lambda could be a parameter that has different value for every node
18 Self messaging
 Need to schedule the packet arrival event.
 Remember that messages are omnetpp’s representation of events.
 We need to send ourselves a message in the future, to trigger the sending at the
appropriate time.
 The scheduleAt(t, msg) function allows us to send ourselves a message that will be
received at simulation time t
 The simTime() functions returns the current simulation time.
 We can combine this with our samples interval to send ourselves a message as
follows:
19 Self messaging cont.

 Then inside the handleMessage(cMessage *msg) function, we can check if this is a

self message or not using the isSelfMessage() function
20 Sending the packet

 Finally, we can use the send(msg, gatename,

gateindex) function to send our message on the
proper gate to our destination node.
 Example:
 Send(msg, “outs”, 1);
21 Putting it all together

 Live overview of the Mesh project.

22 Extra illustration

Node Output gate Connected to

Node 0 0 -> Node 1
1 -> Node 2
2 -> Node 3
Node 1 0 -> Node 0
1 -> Node 2
2 -> Node 3
Node 2 0 -> Node 0
1 -> Node 1
2 -> Node 3
23 Useful functions

 gateSize(“outs”) //inside c, returns the size of the “out” gate array

 getIndex() // inside c, returns the current node index
 getParentModule()->par(“n”) //inside node.c, returns a pointer to the parent
module Mesh, and then we can use this pointer to access any parameter in the
parent module.
 Mesh.n // inside .ini where Mesh is the network module.
 Mesh.nodes[..] .lambda //inside.ini where the [..] means all the array
elements.
24 Requirement
 Similar to the Mesh network shown in the lab, create a Star network
topology consisting of 1 HUB and N nodes.
 N should be configurable in the .ini file.
 Each node is connected only to the hub.
 The hub is connected to all nodes.
 Any node is able to send to any other node through the hub.
 Sample from an exponential distribution to get the inter-packet wait times
using parameter Lambda which is also configurable in the .ini file.
25 Submission
 Work in pairs
 Submit one .zip project folder
 In the private comments write your names and Id’s.
 Submit one short video [ 3 to 5] minutes showing your working project
and showing each file of your project describing what changes you made.