A High Performance and Efficient TCP Variant

Abstract

With emergence of the latest technology and deployment of wireless mobile networks for data communication services, the need of the people for faster and robust paradigm has also been increased. There has been a significant effort to tune TCP for these networks. Various TCP variants have been proposed. The performance of TCP is affected due to several factors including congestion window, maximum packet size; retry limit, recovery mechanism, backup mechanism and mobility. These variants are successful in fixed networks but do not yield good results in a mobile wireless network. In this paper, we propose a new TCP variant named TCP University of Bridgeport (UB) by integrating the features of TCP Westwood and Vegas. TCP-UB provides better performance than the other TCP variants from the mobility point of view. We have simulated our algorithm using NS2.28, which shows that TCP-UB achieves superior performance over TCP Vegas and Westwood. The algorithm yields better results in terms of goodput due to effect of the mobility.

Full text

2012 ASEE Northeast Section Conference University of Massachusetts Lowell
A High Performance and Efficient TCP Variant
1
Wafa Elmannai,
2
Khaled Elleithy,
3
Abdul Razaque
Abstract – With emergence of the latest technology and deployment of wireless mobile networks for data
communication services, the need of the people for faster and robust paradigm has also been increased. There has
been a significant effort to tune TCP for these networks. Various TCP variants have been proposed. The
performance of TCP is affected due to several factors including congestion window, maximum packet size; retry
limit, recovery mechanism, backup mechanism and mobility. These variants are successful in fixed networks but do
not yield good results in a mobile wireless network. In this paper, we propose a new TCP variant named TCP
University of Bridgeport (UB) by integrating the features of TCP Westwood and Vegas. TCP-UB provides better
performance than the other TCP variants from the mobility point of view. We have simulated our algorithm using
NS2.28, which shows that TCP-UB achieves superior performance over TCP Vegas and Westwood. The algorithm
yields better results in terms of goodput due to effect of the mobility.
Keywords: TCP-UB, TCP Westwood, TCP Vegas, Efficiency and Mobility.
INTRODUCTION
The use of wireless networks in this century has motivated many researchers to study and make exertion for
modifying TCP, which was originally designed for fixed wired networks. The result of those efforts shows that TCP
in its current structure is not an optimal transport service provider for mobile and wireless networks. Several TCP
variants have been intended and implemented in order to optimize the performances, which cause packets loss.
However, TCP performs well with these networks by improving the congestion mechanism of the existing TCP
variants [5].
First, TCP Vegas is interoperable variant, which augments the throughput performance, reduces packet loss, while it
does not affect the fairness. The advantage of TCP Vegas is to calculate the available bandwidth in network. TCP
Vegas determines the bandwidth on the basis of difference between expected and actual throughput to avoid packet
loss [2].
Second, TCP Westwood gives more significant improvement in wireless networks with lost links [6]. TCP
Westwood estimates the bandwidth that is employed by the congestion control algorithm. The bandwidth estimation
is performed at the sender side of TCP connection. The slow start and congestion avoidance phases are unaffected
during linear and exponential increase of congestion window.
It has been observed that TCP Vegas performed better than Reno with 4.29% to 9.7%, SACK with 1.64% to 4.66%,
Tahoe with 4.11% to 9.71% and Westwood with 1.12% to 2.9% & New Reno with 2.01% to 5.6%. On basis of
varying mobile nodes and different traffic flows in previous research [10]. Furthermore, the minimum effect of
mobility has been analyzed on TCP Westwood. So, the scope of our research is to introduce a new TCP variant to
provide better performance based on the above results that we obtained.
In this paper we propose a new TCP variant by integrating the important features of TCP Vegas and Westwood to
provide better performance from both efficiency and mobility point of views.
However, in the next Sections we describe the related work, the proposed work, the simulation set up, the simulation
results and analysis, discussion of results and finally conclusion.
__________________________
1
Computer Science and Engineering Department, University of Bridgeport, welmanna@bridgeport.edu
2
Computer Science and Engineering Department, University of Bridgeport, elleithy@bridgeport.edu
3
Computer Science and Engineering Department, University of Bridgeport, arazaque@bridgeport.edu

2012 ASEE Northeast Section Conference University of Massachusetts Lowell
RELATED WORK
In [3] a new policy added to TCP Westwood based Concurrent Multipath Transmission mechanism (CMT). This
policy makes the process of transmission the data and measuring the bandwidth more efficient. The paper provides
the modifying Westwood algorithm (which called CMT-Ww) by adding a new formula to measure the bandwidth
with CWND Update for CMT algorithm (CUC) to improve the algorithm of Westwood. However, the paper is using
SCTCP-CMT module and updating the congestion control mechanism to apply for CMT-Ww. Moreover, the
experiment compared CMT-Reno and CMT-Ww in different paths with setting the bandwidth between each path.
The result could improve the throughput. Also, this experiment could show a good adjusting of CWND based on the
available bandwidth.
[4] Compares TCP variants: TCP Tahoe, Reno, Lite, New Reno, elective Acknowledgement (sack), Westwood,
Vegas and Forward Acknowledgement (Fack) over Mobile Ad-Hoc networks (MANETs) by reviewing the above
variants. The evaluation was specific on different congestion control algorithms such as: congestion avoidance, fast
recovery, slow start, retransmission, fast retransmission, selective acknowledgment and congestion control.
However, the result of this review shows the condition and consumption of the network has an effect on the
behavior of the TCP variants. This effect is based on setting the parameters of these variants. Furthermore, each TCP
variant has its own solution for network problems.
[7] Proposes a new scheme, which is a Threshold –Based Congestion Control Mechanism on TCP Vegas to solve
the progressive performance of Vegas over heterogeneous networks. This research was based on previous research
showed TCP Vegas can perform better than TCP Reno over homogeneous networks but not over heterogeneous
networks. However, the proposed scheme attains higher throughput than the original Vegas without unfairness. By
that, it solves unfairness problem of TCP Vegas for an available bandwidth.
[9] Is based on comparing the performance of both TCP Westwood and TCP Reno over the wireless Ad hoc
networks by using NS2 simulator. The experiment used same topology for the two scenarios with setting the
transport layer between both sides the sender and the receiver. The result of this experiment shows that Westwood
provides a faster recovery mechanism, which can solve the aggressive performance of shrinking. When TCP Reno
reduces the congestion window to the half after the three duplicated acknowledgments arrived.
[6] Analyzes TCP Vegas performance based on several exterminates that already proposed over 4
th
Generation,
Long Term Evolution (LTE) using NS2 simulator. The authors used different values of the two parameters (Alpha,
Beta) over their experiment. This experiment proves that TCP Vegas can perform better than the other TCP variants
when they used different values of the parameters. This provides a great improvement in the throughput.
Some existing TCP Variants over hybrid network are analyzed in [10]. However, the major contribution of this
research is to identify the loss of throughput on different mobility ratios and to design mobility-based hybrid
network with random waypoint mobility model, where TCP Variants are simulated and analyzed. The authors have
shown results that TCP Vegas performs better in Access Point Name (APN) hybrid network where the minimum
effect of mobility was analyzed on TCP Westwood.
However, these papers focused on exiting variants when we compare our new proposed TCP-UB with exiting
algorithm to find out the fairness.
PROPOSED WORK
TCP-UB Algorithms
The main idea of TCP_UB comes up from TCP Vegas and TCP Westwood. This algorithm behaves in the slow start
phase as TCP Vegas exactly and in the congestion avoidance phase behaves as both with adding a new component
called (Gama). However, during the Congestion phase, we automatically adjust three threshold values (Alpha,
Gama, Beta) as is shown in Figure 1.
The congestion window size (CWND) increases by one since the difference of the expected rate and the actual rate
is less than Alpha (a minimal threshold) until it reaches the middle threshold Gama. The reason of that adjustment is
the expected throughput is still low as well as we save the bandwidth. Since the difference is less than Gama (a
middle threshold) then CWND behaves as TCP Westwood with checking the bandwidth each time to know if the
CWND increases or decreases with resetting both the slow start threshold (SSThresh) and the CWND. That

2012 ASEE Northeast Section Conference University of Massachusetts Lowell
continues until it reaches the highest threshold Beta which is (a maximum threshold) then CWND decreases or
keeps constant since the expected throughput gets high.
The scope of this section is to show the strength of our algorithm and the ability of efficient usage of Bandwidth.
This algorithm works as given below:
If (the Dup ACKs are arrived) then
{
Let Base RTT is the minimum of all RTTs; // RTT: Round Trip Time
Expected Rate= CWND /Base RTT; //Base RTT: the minimum RTT
Actual Rate= CWND/RTT; // to estimate the flow throughput
Diff = (Expected Rate – Actual Rate) BaseRTT; // Diff: the difference between the expected and actual rate
If (Diff <) then //: Alpha (Minimum threshold)
CWND+1
else
If (Diff=) then // is a new variable to estimate the congestion possibility (Gama)
{
Let ssthresh = (BWE*Base RTT)/ seg_size; /* BWE: the Bandwidth Estimation; Seg_size: the
size of the segment*/
If (CWND > sthresh) then
CWND=ssthresh
else
If (the time out is expired) then
{
Let CWND=1;
ssthresh = (BWE*BaseRTT)/seg_size;
If (ssthresh<2) then
Ssthresh=2;
}
}
else
If (Diff >β) then // β: Beta (Maximum threshold)
CWND-1;
Otherwise -> CWND;
};

2012 ASEE Northeast Section Conference University of Massachusetts Lowell
Time
SSThresh
128
64
32
16
8
4
2
0
5 10 15 20 25 30 35 40
Alpha
Gama
Beta
Figure1: The Congestion Control Behavior of TCP-UB
SIMULATION SET UP
The scenarios are simulated using ns2.28 on LINUX Red Hat-9. We have implemented TCP-UB algorithm by
Object-Oriented extension of TCL (OTCL). The Implementation of TCP-UB is shown in the Figure 2. We use
1000*1000 squares meters for the simulation area and simulation time is 140 seconds. The transmission range is 250
meter; the nodes cannot transmit after this limit [1]. TCP Westwood and TCP Vegas are simulated and used for
comparison to ensure the efficiency of TCP variants with TCP-UB on wireless and MANET. The packet size is
1040 bytes with 40 bytes payload. Each node can send 8packets/sec. The buffer size of the queue is 80 packets. 100
mobile nodes are placed over the networks. Random Waypoint Model (RWM) is imitated for starting nodes’
location. We set the pause time for 5 seconds for each 50 seconds. The minimum speed of the mobile node is 0
m/sec and the maximum speed is 35m/sec. By dividing both the minimum and the maximum speed of the node, we
can get the moving speed randomly.
Figure2: The implementation of TCP-UB over Wireless and MANET Networks.

2012 ASEE Northeast Section Conference University of Massachusetts Lowell
Furthermore, we use the Dynamic Source Routing (DSR) to obtain a goodput performance for routing over the
network. We use Contain Bit Rate (CBR) and File Transfer Protocol (FTP) applications, which support TCP and
User Datagram Protocol (UDP). Finally, Antenna type is Omni Directional to support the simulation area.
SIMULATION RESULTS AND ANALYSIS
In this section we discuss the simulation scenarios.
A. Efficiency Variance Scenario
In this scenario, we have simulated our network over MANET and wireless segments with NS2 and examined the
efficiency of TCP Westwood, TCP Vegas and TCP-UB. For each of above TCP variants, we have collected their
acknowledged and received packets. In this scenario, the average speed is 17.5 m/sec for each TCP variants with
Random Waypoint Mobility model.
Figure 3 shows the efficiency of TCP Vegas, which steadily decreases for acknowledged packets from 5.8Mb to
4.2Mb over the time. In Figure 4 we can notice that the efficiency of TCP Westwood decreases with almost the
same numbers. The reason for this decreasing of packet’s efficiency is the mobility. This scenario covers MANET
and wireless. In this condition, MANET stays dynamic make radio channel fading and mobility of nodes are main
effects.
The mobile nodes take longer time to recover from broken links. The mobility of nodes has an effect on TCP
variants due to changes of routing information over the network and can cause longer RTT and repeated timeouts.
In fact, the mobility of nodes can make the receiver getting out of order packets which can affect the
acknowledgements. However, that can cause the duplicating acknowledgement and starting retransmission
algorithm with reducing in the congestion window [10].
On the basis of efficiency, it is clear that TCP-UB acknowledges more packets than TCP Vegas and TCP Westwood
as shown in Figure 5. These data shows TCP-UB received and acknowledges more packets compared with other
variants. Furthermore, an important feature of TCP-UB is the stability. The performance of TCP-UB becomes stable
during all the simulation time.
Figure 3: show the performance of TCP-Vegas Figure 4: show the performance of TCP- Westwood

2012 ASEE Northeast Section Conference University of Massachusetts Lowell
Figure5: show the performance of TCP-UB
B. GoodPut Scenario
We show average of goodput for TCP Vegas, TCP Westwood and TCP-UB from static and mobility point of view
as are shown in Figure 6 and 7.
Figure 6: Mobility scenario of Goodput Figure 7: Static Scenario of Goodput
TCP Vegas and Westwood are not stable if the speed increases from 25 to 35m/sec. They show poor performance
while TCP-UB the most stable performance throughout changes in nodes’ speed. The changes in speed do not affect
the performance of TCP-UB because including of Gama, the goodput of TCP- UB is better than other TCP variants.

2012 ASEE Northeast Section Conference University of Massachusetts Lowell
Another important factor is using Gama for division of congestion avoidance phase into three parts. The partition of
congestion avoidance phase provides sufficient time to control congestion window and loss of packets. The behavior
of routing protocols also cannot affect Good put performance of TCP-UB.
The performance of TCP-UB, TCP Westwood and TCP Vegas is shown in Figure 6 for Mobility view and in Figure
7 for static view.
DISCUSSION OF RESULTS
We have simulated TCP-UB in NS-2 over MANET and wireless networks by using DSR routing protocol. Two
types of scenarios are generated: static and mobility. We mainly focus on mobility based scenario to measure the
performance of exiting TCP Vegas, Westwood and our newly TCP-UB. We use varying speed of mobile nodes to
examine the real behavior of variants.
In fact, All of TCP variants are affected due to mobility because they do not have completely mobility aware model.
However, it has been shown that TCP-UB is better than other TCP variants adding extra component Gama. It gives
an excellent result for acknowledged and received packets, and good RTT at different mobile speeds. The reason of
producing better performance is to deploy of new component because window does not shrink many times.
Finally, TCP Vegas and TCP Westwood have been designed to achieve excellent results in fixed wired networks but
they perform less when compared with the TCP-UB over the hybrid network.
CONCLUSION AND FUTURE WORK
Our proposed protocol can be an advantageous system when it is used for Ad Hoc Networks (MANET) in many
applications like military environment, scattered hospitals and industry environments. These applications require
data over mobile nodes. We have used the NS2.28 with Random Waypoint Mobility Model to integrate some
features of both TCP Vegas and TCP Westwood to provide better performance.
We have used MANET and wireless network to achieve better performance. Furthermore, experimental results
show that TCP-UB is not affected by mobility, meanwhile achieves high efficiency and higher delivery of data as
compared to TCP Vegas and TCP Westwood. TCP-UB has achieved excellent performance based on the results we
validated and measured over several conditions and scenarios.
The implementation of this algorithm shows that there are still more challenges related to TCP variants that can be
addressed in the future. We are planning in the future to analyze and evaluate TCP-UB, TCP Vegas and TCP
Westwood in MANET and wireless network with respect to bandwidth consumption and congestion control
scenario.
REFRENCES
[1] Lo Anthony, Zhou Jinglong , Jacobsson Martin, Niemegeers Ignas, “NS-2 Models for Simulating a Novel
Beyond 3G Cellular Multi- hop Network,” ACM New York, USA, 2006, pp.1-4.
[2] Brakmo Lawrence, Peterson Larry, “TCP Vegas: End- to- End Congestion Avoidance on a Global Internet,”
Journal on Selected Areas in Communications, IEEE, 1999, pp. 1465-1480.
[3] Lan Kou, Sha Niu, “A CMT Congestion Window Updates Mechanism Based on TCP Westwood,”
International Conference on Mechatronic Science, Electric Engineering and Computer (MEC), China, 2011, pp.
2119 - 2122.
[4] Tayade Mandakini, Sharma Sanjeev, “Review of Different TCP Variant in AD-hoc Networks,” International
Journal of Engineering Science and Technology (IJEST), Chennai, 2011, pp.1906-1913.
[5] Abed Ghassan, Ismail Mahamod, Jumari Kasmiran, “A Survey on Performance of Congestion Control
Mechanisms for Standard TCP Versions,” Australian Journal of Basic and Applied Sciences, Pakistan, 2011,
pp.1345-1354.
[6] Abed Ghassan, Ismail Mahamod, Jumari Kasmiran, “Characterization and observation of (transmission control
protocol) TCP-Vegas performance with different parameters over (Long term evolution) LTE networks,”
Academic Journals, Scientific Research and Essays, Malaysia, 2011, pp. 2003-2010.

2012 ASEE Northeast Section Conference University of Massachusetts Lowell
[7] Cheng Rung-Shiang, Ke Chun-Yu, “An Threshold-based Congestion Control Mechanism for Vegas TCP over
Heterogeneous Wireless Networks,” 6th International Symposium on Wireless and Pervasive Computing (ISWPC),
Hong Kong, 2011, pp. 1-6.
[8] UCN/LBL/VINT,"Networksimulator-NS2," htp:/ww.isi.edu/nsnani/ns, 2004.
[9] Dobhal Dinesh, Sharma Rakesh, “Simulation Base Analysis of TCP Reno and TCP Westwood Over IEEE
802.11 Wireless Ad hoc Networks,” International Journal of Computer Science & Communication, 2010, pp. 383-
386.
[10] Elmannai Wafa, Razaque Abdul, Elleithy Khaled, "Simulation based Study of TCP Variants in Hybrid
Network," 2011 ASEE Northeast Section Conference, University of Hartford, USA, 2011, p1-6.
Brief Biographic:
Dr. Elleithy
is the Associate Dean for Graduate Studies in the School of Engineering at
the University of Bridgeport. He has research interests are in the areas of network
security, mobile communications, and formal approaches for design and verification. He
has published more than one hundred fifty research papers in international journals and
conferences in his areas of expertise.
Dr. Elleithy is the co-chair of the International Joint
Conferences on Computer, Information, and Systems Sciences, and Engineering
(CISSE). CISSE is the first Engineering/Computing and Systems Research E-Conference
in the world to be completely conducted online in real-time via the internet and was
successfully running for four years.
Dr. Elleithy is the editor or co-editor of 10 books
published by Springer for advances on Innovations and Advanced Techniques in
Systems, Computing Sciences and Software.
Mr. Abdul Razaque is PhD student of computer science and Engineering department in
University of Bridgeport. His current research interests include the design and
development of learning environment to support the learning about heterogamous
domain, collaborative discovery learning and the development of mobile applications to
support mobile collaborative learning (MCL), the congestion mechanism of transmission
of control protocol including various existing variants, delivery of multimedia
applications. He has published over 30 research contributions in refereed conferences,
international journals and books. He has also presented his work more than 10 countries.
During the last two years he has been working as a program committee member in IEEE,
IET, ICCAIE, ICOS, ISIEA and Mosharka International conference. Abdul Razaque is
member of the IEEE, ACM and Springer Abdul Razaque served as Assistant Professor at
federal Directorate of Education, Islamabad. He completed his Bachelor and Master
degree in computer science from university of Sindh in 2002. He obtained another Master
degree with specialization of multimedia and communication (MC) from Mohammed Ali
Jinnah University, Pakistan in 2008. Abdul Razaque has been directly involved in design
and development of mobile applications to support learning environments to meet
pedagogical needs of schools, colleges, universities and various organizations.
Mrs. Wafa Elmannai is a Master student
in the School of Engineering and Computer
Science at the University of Bridgeport. She has finished her Bachelor’s degree in
Fall2005 at Ben Ashore College for Computer Science. She graduated with highest
honors, ranking 4th in her Bachelor degree and awarded merit certificate.
Mrs. Elmannai worked as a Research Assistant at University of Bridgeport since 2010-
2011. She worked in management department as assistant administrator in education
department since 2005-2009. She worked also as teacher in Algamaheria School since
2004-2005. Mrs. Elmannai is interested in network area, mobile communication, and
some software applications. She has published several papers in international
conferences. She mad presentation of her research in conferences.