College of Computing

Department of Computer Science (MSc)

ADVANCED COMPUTER NETWORKS PAPER ON

EXPLORING CONGESTION CONTROL MECHANISM OF TCP
VARIANTS OVER WIRED & WIRELESS NETWORKS:

Course code: - CoSc-6110

Date of submission: 17 March 2023.

Submitted to Course Instructor: - Mebratu Fana (Assistant. Prof)

Madda Walabu University (MWU), Ethiopia

17 March 2023
Table of Contents
1. Introduction .................................................................................................................................. 1
2. Overview of Transmission Control Mechanisms ........................................................................ 1
1. Congestion Control Mechanisms............................................................................................. 1
1.1 Slow start ............................................................................................................................. 1
2.1 TCP’s Congestion Avoidance .............................................................................................. 1
3.1 TCP’s Fast Retransmission .................................................................................................. 2
4.1 Fast Recovery....................................................................................................................... 2
3. Experimental Results and Discussions ........................................................................................ 2
4. TCP Congestion Control Algorithms in Wired Network ............................................................ 2
1. Comparison of Congestion window concerning time ............................................................. 3
2. Comparison of throughput (kbps) with respect to time (Second) ........................................... 3
3. Numerical Analysis ................................................................................................................. 3
5. Evaluation of TCP Congestion Control Algorithms in Wireless Networks ................................ 4
1. Comparison of Congestion window concerning time ............................................................. 4
2. Comparison of throughput (kbps) with respect to time (Second) ........................................... 5
1.1 Comparison of Throughput In case of AODV routing protocol .......................................... 5
2.1 Comparison of Throughput In case of DSDV routing protocol .......................................... 5
3. Numerical analysis of TCP over MANET .............................................................................. 6
6. Critique and My Argument .......................................................................................................... 6
7. Conclusion ................................................................................................................................... 7
8. Preference ...................................................................................Error! Bookmark not defined.

Table 1:- Comparison of Simulated data for TCP variants is using different Parameters in Network
Topologies. ....................................................................................................................................................... 4
Table 2:- Simulated Data Table of Network Topology-1, 2,3 & 4 using TCP variants using AODV ............. 5
Table 3:- Simulated Data Table of Network Topology-1,2,3&4 using TCP variants in case of DSDV .......... 5

I
ABSTRACT
A reliable end to end communication is a bulky concept which is promised by the transport layer protocol
TCP. TCP, a Reliable transport protocols are functioning to perform well in different networks but, packet
losses occur mostly because of congestion. TCP contains several mechanisms (such as slow start,
congestion avoidance, fast retransmit and fast recovery) for ensuring reliability. However, it has reached its
limitation in some challenging network environments like-High speed communication, Communication over
different media. Thus, it requires further analysis and development of congestion control algorithms. In this
paper, I have explored the reliability and robustness of TCP variants (Tahoe, Reno, New-Reno, SACK,
FACK and TCP VEGAS, HSTCP, CUBIC TCP) based on different parameters such as throughput,
end-to-delay, jitter, and packet drop ratio over wired and wireless networks. Based on a review paper I
have also compared and discussed different congestion control and avoidance mechanisms of TCP variants
to show how they affect network parameters and put my argument on this paper review.
Index Terms: - Congestion avoidance, Congestion Window, Fast-recovery, Fast-retransmit, Reliability,
Slow Start, TCP variants.

II
 1. Introduction
TCP (Transmission Control Protocol) is mostly used Internet Protocol. The main feature of TCP is
that it can handle the congestion. When the packet sending rate is increased than the receiving rate
then congestion arises. Traditional TCP implementation did little to minimize network congestion.
Modern TCP implementations contain Additive Increase or Multiplicative Decrease [1]with four
intertwined algorithms (slow start, congestion avoidance, fast retransmit and fast recovery) aimed
at controlling network congestion while maintaining good user parameters like throughput. It
ensures reliability by starting a timer whenever it sends a segment and the receiver that
acknowledge the segments that it receives. If it does not receive any acknowledgment from the
receiver within the ‘time-out’ period, it retransmits the segment again. It is very important to
retransmit a packet to alleviate common network problem in different network environments that is
why I agree with this paper review. I reviewed the paper by taking a brief look at different
congestion avoidance algorithms and noting how they differ from each other on this crucial idea
of the paper. In this paper, I try to represent a performance compare to clarify the main differences
among the TCP variations further.

2. Overview of Transmission Control Mechanisms

1. Congestion Control Mechanisms
1.1 Slow start
Slow-start is a mechanism used to gradually increase the amount of data in transmission and attempts to
keep the segment uniformly spaced. It is one of the most critical parts of the congestion avoidance
technique used by TCP as specified by RFC 5681 [2]. This technique is accomplished in conjunction with
other algorithms to avoid sending more data than the network is capable of transmitting, to avoid network
causing congestion. In the slow start, when a connection is established, first the value of cwnd is set to 1,
and after each received ACK the value is updated to (cwnd = cwnd + 1), i.e., double of cwnd for each RTT.
The exponential growth of cwnd continues until a packet loss is observed, causing the value of ssthresh to
be updated to (ssthresh = cwnd/2). After the packet loss, the connection starts from the slow start again
with cwnd = 1, and the window is increased exponential until it equals ssthresh, the estimate for the
available bandwidth in the network. At this point, the connection goes to the congestion avoidance phase
where the value of cwnd is increased less aggressively with the pattern (cwnd = cwnd + 1/cwnd), implying
linear instead of exponential growth. This linear increase will continue until a packet loss is detected. That is
why it is also known as the exponential growth phase.
2.1 TCP’s Congestion Avoidance
It is one of the most critical factors that determine the number of bytes that can be outstanding at any time.
Congestion usually occurs when data arrives faster and sends out at a lower speed [3]. The sender maintains
the congestion window where the size of this window is calculated by estimating how much congestion
there is between the two places. The basis of TCP congestion control mechanism lies in Additive Increase
Multiplicative Decrease (AIMD), halving the congestion window for every window containing a packet
loss, and increasing the congestion window by roughly one segment per round-trip time/delay or
RTT otherwise. If all segments are received, and the acknowledgments reach to the sender on time,
some constant is added to the window size.
The window keeps growing exponentially until a timeout occurs or the receiver reaches its limit (a

Congestion Control Mechanism Of TCP Variants Over Networks Paper Review page 1
threshold value "thresh"). After this, the congestion window increases linearly at the rate of 1/
(congestion window) packets on each new acknowledgment received. When a packet is dropped,
then the congestion window (W) reduced to half. In this section [4] a steady-state model, is used in
the development of TCP’s congestion window graphically to depict how packet is dropped.
3.1 TCP’s Fast Retransmission
If an acknowledgment is not received for a particular segment within a specified time (Round-trip
delay time), the sender will assume the segment was lost in the network and will retransmit the
segment. The fast-retransmit works as follows: if a TCP sender receives four acknowledgments
with same acknowledge number then, there is enough evidence of packet drop with the higher
sequence number. The sender will retransmit the packet before its timeout. It means that instead of
waiting for the retransmit timer to expire, the sender can retransmit a packet immediately after
receiving three duplicate ACKs.
4.1 Fast Recovery
The Fast-Recovery algorithm is implemented together with a Fast-Retransmit algorithm that
retransmits the missing packet signaled by three duplicate ACKs and wait for an acknowledgment
of the entire transmit window. It is also called Fast retransmit/Fast-Recovery algorithm. The fast
recovery is an improved version of fast retransmit and algorithms are usually implemented together
[5] as follows:
a) When the third duplicate ACK in a row is received, set ssthresh to value:
Equation 1: ssthresh = min (cwnd/2, 2 MSS) ------------------------------- (1)
where MSS=maximum segment size.
Retransmit the missing segment. Set cwnd to ssthresh plus three times the segment size.
This increases the congestion window by the number of segments that have left the network and
which the other end has cached.
b) Each time another similar ACK arrives, which increase cwnd by the segment size. This also
inflates the congestion window for the additional segment that has left the network. Transmit
a packet, if allowed by the new value of cwnd:
Equation 2: cwnd = (ssthresh + no. of duplicate acks received) ------ (2)
c) When the next ACK arrives, that acknowledges new data packet, set cwnd to ssthresh. This
ACK is the acknowledgment of the retransmission of data from step 1, one round-trip time after
the retransmission. Additionally, this ACK should acknowledge all the intermediate segments
sent between the lost packet and the receipt of the first duplicate ACK. This step is congestion
avoidance since TCP is down to one-half the rate it was at when the packet was lost.
 NB:- This paper is done using a quantitative paper review data collection
measurement, As most graphs are drawn, tables and Network simulation are used
to make an algorithms work successfully.

3. Experimental Results and Discussions

4. TCP Congestion Control Algorithms in Wired Network
In this section, my review paper performed the simulation to investigate the performance of the
various congestion control algorithms (Taho, Reno, Newreno, Sack, Fack and Vegas, HSTCP,
Cubic) in TCP from different aspects in a wired network. Previous work [6] [7] [8]performs

Congestion Control Mechanism Of TCP Variants Over Networks Paper Review page 2
comparison considering very few parameters. In this comparative analysis, I tried to investigate
most of the parameter to explore actual behavior of TCP protocol over wired and wireless network.
 Here we use network simulator version-2.35 [9] [10]where TCL and OTCL scripting
[11]are used for better scheduling event and controlled environment.
1. Comparison of Congestion window concerning time
Congestion window (CWND) of TCP changes based on the change of its basic algorithms in every
TCP variant (Taho, Reno, Newreno, Sack, Fack, Vegas, HSTCP, Cubic TCP). Simulation result of
congestion window describes slow start, congestion avoidance, fast retransmit and fast recovery
algorithms in TCP variants. Graphically the overall comparison of all TCP variant for low and high
bandwidth network in the congestion window versus time graph is scratched.
The result collected from an experiment/data survey from the comparison is, network traffic
prediction, speed and congestion control mechanism. Based on the experiments and
comparison done on high speed networks, I concluded that CUBIC TCP variant shows its
highest performance in the high-speed network where cubic TCP is best but in the low-speed
network, they are completely inefficient.
2. Comparison of throughput (kbps) with respect to time (Second)
The throughput of TCP packets and Acknowledgement changes based on the change of its basic
algorithms in every TCP variant. Like above experiment I comparison of throughput vs. time for
TCP Taho, Reno, Newreno, Sack, Fack, Vegas, HSTCP, and Cubic TCP was made in the paper.
As it can seen, initially the throughput and acknowledgment value increase abruptly and then its
remain constant with respect to time for rest of the period, which indicates that packets delivery per
RTT is a constant, i.e., the same number of packets are delivered by the network in a certain
amount of cyclic period.
The last & most modern variants, HSTCP & Cubic TCP show the lowest throughput value for
the given topology, as it starts from slow start phase every time after retransmission than other
in graph comparison. Finally I concluded that CUBIC TCP shows its very efficient
performance in the high-speed network, but TCP Vegas is entirely in inefficient in challenging
network situations.
3. Numerical Analysis
The throughput of TCP packets and Acknowledgement changes based on the change of its basic
algorithms in every TCP variant. From graphical representation shown the comparison of
throughput vs. time for TCP Taho, Reno, Newreno, Sack, Fack, Vegas, HSTCP, and Cubic TCP.
As we can see, initially the network parameters comparison is done on these all TCP Variants.

Congestion Control Mechanism Of TCP Variants Over Networks Paper Review page 3
Table 1:- Comparison of Simulated data for TCP variants is using different Parameters in Network Topologies.

5. Evaluation of TCP Congestion Control Algorithms in Wireless Networks

When there is no infrastructure in the network, the mobile ad hoc network is the best choice. We
need to analyses the routing mechanisms which are called the MANET routing protocols. In this
section, paper perform simulation in mobile ad-hoc networks to investigate the performance of the
various congestion control algorithm (Tahoe, Reno, Newreno, Sack, Fack, Vegas, HSTCP, Cubic)
in TCP from different aspects over two major routing protocols AODV and DSDV. In this analysis,
the simulations are performed with four slightly different simulation scenarios that contain 3, 20,
30, 45 nodes respectively.
1. Comparison of Congestion window concerning time

Currently, all implementations of the TCP congestion algorithm assume that congestion causes
timeouts but, also transmission errors. This study stands when applied to wired networks as they
are relatively reliable and show very low errors. However, this concept does not stand for wireless
network as they suffer from high error and packet loss rates.
Congestion window (CWND) of TCP in wireless network changes based on the change of its basic
algorithms in every TCP variant (Tahoe, Reno, Newreno, Sack, Fack, Vegas, HSTCP, Cubic).
Simulation result of congestion window describes slow start, congestion avoidance, fast retransmit
and fast recovery algorithms in TCP variants.

Graphically an overall comparison of the congestion window versus time among all the TCP
variants in different network situation discussed from the paper with in their node respectively.
I concluded based on congestion control, that in case of small number of nodes, TCP Taho, Reno,
Newreno gives highest performance and in network scenarios where the number of nodes more
significant than 30, CUBIC TCP is best suited when AODV routing protocol is used, and TCP
Newreno is best suited when DSDV routing protocol is used as it can change its congestion
Congestion Control Mechanism Of TCP Variants Over Networks Paper Review page 4
window based on network traffic situation but in high bandwidth network situation it completely
inefficient. Again, according to the figure above we can also conclude that Cubic & high-speed
TCP shows its highest performance in network scenarios where the number of the node is higher
than 30 and the routing protocol is AODV.
2. Comparison of throughput (kbps) with respect to time (Second)
1.1 Comparison of Throughput In case of AODV routing protocol
The comparison of throughput and versus time for TCP Tahoe, Reno, Newreno, Sack, Fack, Vegas,
HCTCP, CUBIC in case AODV routing protocol is made on simulation which shows quite a
different result from above comparison on different nodes.
So finally, I concluded that CUBIC TCP shows its very efficient performance in the network when
the number of the node is higher than 30 but TCP Tahoe, Reno, Newreno gives an efficient
performance when some nodes are less than 10 and is completely inefficient in challenging
network situation.

Table 2:- Simulated Data Table of Network Topology-1, 2,3 & 4 using TCP variants using AODV

2.1 Comparison of Throughput In case of DSDV routing protocol

For a given wireless network topology-4 because of its wise changes in slow start, congestion
avoidance, and retransmission algorithms. The throughput of TCP Reno is also best than Newreno,
Tahoe, Sack, Fack and improves performance when the number of nodes is less than or equal to 20.
So finally, I concluded that TCP Newreno shows it is very efficient performance in the network
where the number of the node is 45 and the routing protocol is DSDV but TCP Tahoe, Reno is
efficient in network situation where the number of the node is less than 10.

Table 3:- Simulated Data Table of Network Topology-1,2,3&4 using TCP variants in case of DSDV

Congestion Control Mechanism Of TCP Variants Over Networks Paper Review page 5
 3. Numerical analysis of TCP over MANET
At the end simulation for each TCP variants, the performance parameter was calculated which will further
help to identify which TCP version is best suited for which situation. From the table drawn in the review
paper I summarized that CUBIC TCP is best suited for a network where the number of the node is higher
than 45 and routing protocol is AODV and TCP Newreno is best for a network where the number of the
node is higher than 45 and Routing protocol is DSDV. Again, TCP Reno performs its best when in both of
AODV and DSDV routing protocol when the number of the node is less than or equal to 20 nodes.

6. Critique and My Argument

Most probable I support this idea of exploring TCP variant on wired and wireless networks to decrease the
impact of network congestion and its parameters. But the author of this paper must again rewrite some
missing concepts on this network congestion algorithm regarding about TCP variants and there usage..
At the beginning of the paper I am confused here to identify the how exponential growth is better than linear
growth on detecting a packet loss in slow start algorithm. There is no measurement concept on this slow
start point about the exponential growth and linear growth to detect packet loss.
In my point of view to minimize the time complexity and packet loss the exponential growth detection is
very important than linear.
The author of this paper must explain about four acknowledgments to make vivid about
retransmission techniques to minimize packet losses. What are a four acknowledgment in this
algorithm?
 There are equations that are used in this paper at the beginning of congestion controlling
mechanism; I don’t know the functionalities of writing this equation because they are not
used in this review paper at all.

Actually, all TCP variants work same, but they are used different mechanism in congestion
handling. Here I can say that the performances of TCP variants are affected by its congestion
control mechanism these needs the further challenges on this paper in the future.
Finally I will give recommend this paper to user network simulation in a vivid manner, drawing a
node, posting a code to minimize the packer send, received ,throughput ,delay and add the amount
of energy level the resource consume, because these are the common network problem we face in
TCP Congestion control. I will not have words because the ideas of this paper very important.

Congestion Control Mechanism Of TCP Variants Over Networks Paper Review page 6
 7. Conclusion
In this paper work, the analysis is done on TCP variants named TCP Tahoe, Reno, Newreno, Sack,
Fack, Vegas, HSTCP, and Cubic TCP in both wired and wireless network. Here I have evaluated
the performance characteristics (Congestion window, Throughput, Delay, Jitter, Packet delivery
ratio, End-to-end delay) of various TCP congestion control schemes under the wired network
conditions with bottleneck end-to-end link capacities. I tried to find some critical cases in which
TCP Reno, TCP New Reno, TCP Vegas, Cubic TCP, HSTCP make some performance
improvements compared to all simulated TCP variants. From simulated data and graphs obtained,
I tried to find which TCP variants are better for challenging network situations from the paper. In
low bandwidth network, I find that both TCP Vegas and TCP SACK make some performance
improvements to TCP Reno. TCP Vegas achieves higher throughput than Reno and SACK for
large loss rate. TCP SACK is better when more than one packet dropped in one window. TCP
Vegas causes much fewer packets retransmissions than TCP Reno and SACK. I have also
identified that TCP Vegas does lead to a fair allocation of bandwidth for different delay
connections. Both TCP Reno and SACK bias against long delay connections. But TCP Vegas
shows very low throughput in case of high bandwidth network. In high bandwidth network, Cubic
TCP achieves higher throughput compared to all other TCP variants. But one drawback of Cubic
TCP is its low throughput in low bandwidth network. I have also concentrated on the behavior of
TCP’s reliability in the mobile ad-hoc network. In MANET, TCP Newreno, Cubic TCP achieve the
highest throughput when the routing protocol is DSDV and AODV respectively, and a number of
the node is higher than 30. Finally, I have put my argument and some suggestion in improving the
drawback of TCP variants and try to comment which variation is better for which network situation
and network parameters identification is easily found.

Congestion Control Mechanism Of TCP Variants Over Networks Paper Review page 7
 8. Bibliography
[1] J. (. a. Y. Zhen., "“Comparison of different TCP congestion control mechanism," Linear Networks and Systems.
Belmont, Calif.: Wadsworth, pp, no. Link: http://www2.ensc.sfu.ca, , pp. 123-135, 1993.
[2] V. P. E. B. M. Allman, "“TCP Congestion Control”,," RFC5681, September 2009.
[3] P. P. S. G. a. V. N. P. Digvijaysinh B Kumpavat, "“Comparison of TCP congestion control mechanisms
Tahoe,Newreno and Vegas”.," IOSR Journal of Electronics and Communication Engineering, vol. Volume 5, no.
Issue 6, pp. PP 74-79.
[4] S. Floyd., "“Promoting the Use of End-to-End Congestion Control in the Internet”," IEEE/ACM Transactions on
Networking,, vol. Volume 7, no. Issue 4.
[5] W. Stevens., "“TCP Slow Start, Congestion Avoidance, Fast Retransmit, and Fast Recovery Algorithms”," RFC
2001, January 1997.
[6] Md. Shohidul Islam, M.A Kashem, W.H sadid, M.A Rahman, M.N Islam, S. Anam., "“TCP Variants and
Network Parameters: A Com prehensive Performance Analysis," nternational Multiconference of Engineerings
and Computer Scientists 2009, pp. pp:351-353, March 18-20 2009.
[7] L. S. B. a. L. L. Peterson., "“TCP Vegas: End to End Congestion Avoidance on a Global Internet”," IEEE Journal
on Selected Areas in Communication, Vols. Volume 13,No 8, pp. pp:1465-1480.
[8] H. (. Z. a. Z. Bian, "Evaluation of different TCP congestion control algorithms using NS-2,,"
Link:http://www2.ensc.sfu.ca/~ljilja/ENSC835/Spring02/Projects/bian_zhan, Date: 5-07-2017.
[9] "The Network Simulator - ns-2, Link: http://www.isi.edu/nsnam/ns,," 7.06.2017.
[10] Paul Meeneghan and Declan Delaney, "An Introduction to NS, Nam,nd Otcl scripting, Department of Computer
Science, National University of Ireland, Maynooth," 2004.
[11] "TCL and OTCL programming,Link:http://www.mathcs.emory.edu/~cheung/Courses/558/Syllabus/A2-Tcl/Tcl-1-
vars.html," Access Date: 6.04.2017.

Congestion Control Mechanism Of TCP Variants Over Networks Paper Review page 8