Description
In this assignment, we are going to use the NS-3 ( www.nsnam.org) simulator, which is a
discrete event, packet level network simulator for Internet systems.
Part 1: 35 points
In this part, we are going to analyze different congestion control protocols. The goal is to
study how the congestion window at the sender varies with time for different protocols.
Create a simple topology of two nodes N1 and N2, and join them using a point-to-point link
as shown in Figure 1. Create a TCP source at N1 and a TCP sink at N2. Fix the application
data rate to be 1 Mbps. The data rate of the link between N1 and N2 is 8 Mbps and
propagation delay is 3ms. At N2 use RateErrorModel as the error model with error rate as
0.00001. Packet size is 3000 bytes.
Figure 1
Connection will start at t=1s and end at t=30s. In our study, we will analyse the following
protocols.
a. Newreno
b. Highspeed
c. Veno
d. Vegas
(Note: You are not required to code these protocols from scratch, use the inbuilt libraries
present in ns-3). Read the usage section from this link on how to change congestion control
protocol in ns3.
1. For each protocol, generate a plot having Congestion window size on the y-axis and
time on the x-axis( till t=30s). (Note: You need to plot 4 separate graphs)
2. For each protocol, find the number of dropped packets in total. What is your
inference?
3. For each protocol, describe what you observed (4-5 sentences per protocol is
enough). You can talk about the trend you observed in the above plot, the algorithms
they used for different phases etc.
[Tip : Before running any file you need to compile/build that file. ns3 uses wscript, to list the
name of all the files that need to be compiled during the building process. Make sure you are
putting right entry in these files before using your custom file. ]
Part 2: 15 points
In this part, we are going to analyze the effect of the bandwidth/application data rate on the
congestion window at the sender. We’ll use the default TcpNewReno as the congestion
protocol.
Create a simple topology of two nodes N1 and N2, and join them using a point-to-point link
as shown in Figure 2. Create a TCP source at N1 and a TCP sink at N2. The channel delay
is 3ms. At N2 use RateErrorModel as the error model with error rate as 0.00001. The
connection will start at t=1s and ends at t=30s. Packet size is 3000 bytes.
Figure 2
a) Plot the congestion window size vs time graph for the TCP connection at different
Channel Data Rates (2Mbps, 4Mbps, 10 Mbps, 20Mbps, 50 Mbps) between N1
and N2. Use Application data rate as 2Mbps. You need to create a plot for each
Channel Data Rate. Explain the trends that you observe. (Note: You need to plot 5
separate graphs)
b) Plot the congestion window size vs time graph for the TCP connection at different
Application Data Rates (0.5 Mbps, 1Mbps, 2Mbps, 4Mbps, 10 Mbps). Use Channel
data rate between N1 and N2 as 6 Mbps. You need to create a plot for each
Application data rate. Explain the trends that you observe. (Note: You need to plot 5
separate graphs)
Part 3: 50 points
In this part, we will learn how to create our own congestion control protocol.
NS-3 by default uses TcpNewReno as the congestion control strategy (Link). It consists of
the following functions of interest to us
1. TcpNewReno::SlowStart:
“ns-allinone-3.29/ns-3.29/src/internet/model/tcp-congestion-ops.cc”
2. TcpNewReno::CongestionAvoidance:
“ns-allinone-3.29/ns-3.29/src/internet/model/tcp-congestion-ops.cc”
Our goal will be to create a new congestion control algorithm called TcpNewRenoCSE. For
this, you need to create 2 new files TcpNewRenoCSE.cc and TcpNewRenoCSE.h in
“ns-allinone-3.29/ns-3.29/src/internet/model”. You can refer to the following link on how to
write a new congestion control algorithm in NS-3.
You need to reuse most of the code of TcpNewReno from
“ns-allinone-3.29/ns-3.29/src/internet/model/tcp-congestion-ops.cc” and
“”ns-allinone-3.29/ns-3.29/src/internet/model/tcp-congestion-ops.h” and the algorithmic
changes need to be done only for the SlowStart and CongestionAvoidance functions in your
TcpNewRenoCSE.cc.
The way you implement it by using inheritance of base class TcpCongestionOps /
TcpNewReno /or any other way is your decision. The important point is that all functionalities
in TcpNewRenoCSE except the below two should be of TcpNewReno.
1. TcpNewReno::SlowStart
2. TcpNewReno::CongestionAvoidance
In Slow Start phase of TcpNewRenoCSE, increase the congestion window size𝐶𝑤𝑛𝑑 as:
𝐶𝑤𝑛𝑑 = 𝐶𝑤𝑛𝑑 + (𝑆𝑒𝑔𝑚𝑒𝑛𝑡𝑆𝑖𝑧𝑒)
1.9
/𝐶𝑤𝑛𝑑
In the Congestion Avoidance phase of TcpNewRenoCSE, update the congestion window
as
𝐶𝑤𝑛𝑑 = 𝐶𝑤𝑛𝑑 + 0. 5 * 𝑆𝑒𝑔𝑚𝑒𝑛𝑡𝑆𝑖𝑧𝑒
In order to understand the behavior of the network using the newly designed
congestion control protocol, we will use the topology described in Figure 3 below.
Figure 3
In Figure.3, N2-N3 and N1-N3 are connected via a point-to-point link. DR means
Data Rate of the channel. Here we have TCP source at N1 and N2 and TCP sink at
N3.
We will use the below 3 configurations of the network described in Fig.3 to analyze
the behavior of congestion control algorithms:
● Configuration1: All senders use TCPNewReno.
● Configuration2: Connection3 at TCP source uses the new congestion
protocol TCPNewRenoCSE. The remaining senders use the default protocol
TCPNewReno.
● Configuration3: All senders use TCPNewRenoCSE.
For all three configurations above, the following hold:
i) Application data rate: 1.5 Mbps
ii) Connection 1 starts at time 1sec, Connection 2 starts at time 5sec,
Connection 3 starts at time 15sec.
iii)Connection 1 ends at time 20sec, Connection 2 ends at time 25sec,
Connection 3 ends at time 30sec.
iv) Packet size is 3000 bytes.
You need to answer the following questions:
For each configuration do the following at each TCP source.
1. Plot Congestion window size vs time (from t=1 to t=30 seconds)
2. Analyze the number of dropped packets for each connection separately.
3. How does the congestion avoidance phase vary on the same sender when using
TCPNewRenoCSE vs TCPNewReno? Explain the observed trends. How does it
impact the entire network?
What to Submit
1. For each part, write your answers, observations, plots and hypotheses in a single
report file.
2. Submission folder should be a single .zip or .tar.gz file containing a report file and
code for each question. Follow the directory structure as given below.
EnrollmentNumber/-
Report.pdf
Q1/-
First.cc (In case more than 1 file then make it
First_1.cc, First_2.cc …)
Q2/-
Second.cc (In case more than 1 file then make it
Second_1.cc,Second_2.cc …))
Q3/-
Third.cc (In case more than 1 file then make it
Third_1.cc,Third_.cc ….))
Congestion/-
TcpNewRenoCSE.cc
TcpNewRenoCSE.h
In case you are using any other helper code/libraries to generate plots, submit
that code for generating plots.
3. The plots can be generated either using the .cc files or any other tool you wish to
choose such as Python. But make sure, that plot is generated just using commands
and not by copy pasting data values manually in excel or something like that during
the demo. (Tip : Save the output values in a separate file and write a script to
process that file and generate plots).
4. For each code file/package/libraries you submitted in the assignment, clearly
describe their purpose in the report (1 to 2 sentences are enough for each).
