EVALUATION OF VIRTUAL PRIVATE NETWORK

IMPACT ON NETWORK PERFORMANCE

By

MUKATSHUNG CLAUDE NAWEJ

Submitted in accordance with the requirements for

the degree of

MAGISTER TECHNOLOGIAE
Electrical Engineering (Computer systems)

at the

UNIVERSITY OF SOUTH AFRICA

SUPERVISOR: PROF SHENGZHI DU

CO-SUPERVISOR: PROF FRANCOIS MULENGA

September 2016
 Abstract

The aim of the study is to investigate what impact the use of VPN has on network
performance. An empirical investigation using quantitative research methods was
carried out. Two sample scenarios were involved in the study: scenario without
VPN and scenario with VPN. In both scenarios, three applications were used in
turns, an HTTP, an FTP, and a CBR. FTP was configured to use window size and
packet size, while CBR used connection rate and packet size. On the other side,
the number of connection was the only parameter used for HTTP. These
applications were injected in a 100 Mbps fixed link of an NS2 simulation
environment. Throughput and delay averages were measured respectively for the
two scenarios and values compared using Student’s t-test.

While the TCP and HTTP throughputs were found decreasing, the UDP
throughput was not affected by the presence of this VPN. Concerning the delay;
the TCP, UDP and HTTP delay were found increasing.

Key words:
Simulation, VPN, NS2, Network performance

1
 Declaration

Student number: 5085-409-7

I declare that the EVALUATION OF VIRTUAL PRIVATE NETWORK

IMPACT ON NETWORK PERFORMANCE is my own work and that all the
sources that I have used or quoted have been indicated and acknowledged by
means of complete references.
I further declare that I have not previously submitted this work, or part of it, for
examination at UNISA for another qualification or at any other higher education
institution.

--------------------------------------- ------------------------------------

Signature Date
(Mukatshung Claude Nawej)

2
 Acknowledgements

Firstly, I would like to thank the One and only one pillar of my strength, God, the
eternal rock of ages for giving me the strength to complete this dissertation.
Secondly, I thank my supervisors, Professors Shengzhi Du and Francois Mulenga
whose help and stimulating suggestions guided me during my research and write-
up of this dissertation.

Next, special thanks go to my beloved kids Eden Nawej and Dan Nawej for
enduring sacrifices which led to this achievement. Special acknowledgements to
my family for their constant love, prayers and appreciation to my desire to go for
further studies.

Finally, I would like to show my deepest gratitude towards my old and childhood
friend Professor Lucks Lukanda Kalobo for his constant encouragement,
motivation and help towards building my confidence.

3
Table of Contents
Abstract ............................................................................................................................... 1
Declaration .......................................................................................................................... 2
Acknowledgements ............................................................................................................. 3
List of Figures ..................................................................................................................... 7
List of Tables ...................................................................................................................... 8
List of abbreviations and acronyms .................................................................................... 9
Chapter 1: Introduction ..................................................................................................... 10
1.1 Background and motivation .................................................................................... 10
1.2 Problem statement ................................................................................................... 10
1.3 Research objectives ................................................................................................. 11
1.4 Research Questions ................................................................................................. 11
1.5 Scope of the study ................................................................................................... 11
1.6 Layout of the dissertation........................................................................................ 12
Chapter 2: Literature review ............................................................................................. 13
2.1 Introduction ............................................................................................................. 13
2.2 Clarification of concepts ......................................................................................... 13
2.3 Open system interconnection .................................................................................. 14
2.3.1 Application layer .............................................................................................. 15
2.3.2 Transport layer ................................................................................................. 15
2.4 Network performance ............................................................................................. 16
2.5 Review of previous researches done on VPN impact ............................................. 17
2.5.1 The impacting aspect review............................................................................ 17
2.5.2 The impacted aspect review ............................................................................. 18
2.6 Summarised findings .............................................................................................. 19
Chapter 3: Research methodology, procedures and techniques ........................................ 21
3.1 Network simulations ............................................................................................... 21
3.2 Simulation Setup ..................................................................................................... 24
3.3 Simulation Procedures ............................................................................................ 26
3.3.1 Procedure case scenario 1 ................................................................................ 26
3.3.2 Procedure case scenario 2 ................................................................................ 28
3.4 Simulation schedule and parameters ....................................................................... 29
3.4.1 Schedule ........................................................................................................... 29

4
 3.4.2 Simulation parameters...................................................................................... 30
3.5 Data collection ........................................................................................................ 31
3.6 Statistical procedures and techniques ..................................................................... 32
3.7 Hardware specifications .......................................................................................... 32
3.8 Software specifications ........................................................................................... 33
3.9 Challenges and difficulties encountered ................................................................. 33
Chapter 4: Effects of VPN on CBR application................................................................ 34
4.1 Introduction ............................................................................................................. 34
4.2 Collected results for CBR ....................................................................................... 34
4.4 Performance evaluation .......................................................................................... 35
4.4.1 Throughput ....................................................................................................... 35
4.4.2 Delay ................................................................................................................ 37
4.5 Summarised findings .............................................................................................. 39
Chapter 5: Effect of VPN on FTP application .................................................................. 40
5.1 Introduction ............................................................................................................. 40
5.2 Simulation results for FTP ...................................................................................... 40
5.4 Throughput’s performance evaluation .................................................................... 41
5.4.1 Window size 10 Kb .......................................................................................... 41
5.4.2 Window size 50................................................................................................ 44
5.4.3 Window size 100.............................................................................................. 45
5.5 Delay’s performance evaluation ............................................................................. 46
5.6 Summarised findings .............................................................................................. 46
Chapter 6: Effects of VPN on HTTP application.............................................................. 47
6.1 Introduction ............................................................................................................. 47
6.2 Simulation results for HTTP ................................................................................... 47
6.4 Performance evaluation .......................................................................................... 48
6.4.1 Throughput ....................................................................................................... 48
6.4.2 Delay ................................................................................................................ 51
6.5 Summarised findings .............................................................................................. 52
Chapter 7: Conclusion and recommendations .................................................................. 53
7.1 Introduction ............................................................................................................. 53
7.2 Summary and conclusion ........................................................................................ 53
7.3 Recommendations and future work ........................................................................ 55

5
References ......................................................................................................................... 55
Appendices........................................................................................................................ 59
Appendix A: Tcl script for scenario 1........................................................................... 59
Appendix B: Tcl script for scenario 2 ........................................................................... 65
Appendix C: Awk script for delay ................................................................................ 71
Appendix D: HTTP throughput .................................................................................... 73
Appendix E: Instant CBR throughput ........................................................................... 80
Appendix F: Instant FTP throughput ............................................................................ 81
APPENDIX F: Instant FTP throughput ........................................................................ 81

6
List of Figures

Figure 2.1: Basic architecture of OSI vs. TCP/IP...…………………..………… 14

Figure 3.1: NS2 command line interface ………………………………………. 21
Figure 3.2: Vim command to create or open tcl script ……………………...…. 22
Figure 3.3: Sample commands for tcl script …………………………………… 22
Figure 3.4: Scenario 1: Network without VPN ………………………………... 23
Figure 3.5: Scenario 2: Network with VPN …………………………………… 23
Figure 3.6: Research flowchart ………..……………………….……………… 24
Figure 3.7: CBR application over UDP in scenario without VPN ……………. 25
Figure 3.8: FTP application over TCPP in scenario without VPN ……………. 26
Figure 3.9: PackMimeHTTP Architecture in scenario without VPN …………. 26
Figure 3.10: CBR application over UDP in scenario with VPN ...……………..27
Figure 3.11: FTP application over TCP in scenario with VPN ………………...27
Figure 3.12: PackMimeHTTP Architecture in scenario with VPN ….……..…. 28
Figure 3.13: Default NS2 trace file ……………………………………………. 30
Figure 3.14: Customised trace file for TCP throughput ……………………….. 30
Figure 4.1: Average CBR throughput ………………………...……………….. 35
Figure 4.2: Average CBR delay of transfer rate 1000 ……………….………... 36
Figure 5.1: Average FTP throughput of window size 10 ……………………… 40
Figure 5.2: Average FTP throughput of window size 50 ……………………… 43
Figure 5.3: Average FTP throughput of window size 100 …………………….. 44
Figure 5.4: Average FTP delay ……………………………………...…………. 45
Figure 6.1: Average Http throughput ………………………………………….. 47
Figure 6.2: Comparison of instant throughput, connection rate 5 ……………... 48
Figure 6.3: Comparison of instant throughput, connection rate 10 ……………. 49
Figure 6.4: Comparison of instant throughput, connection rate 15 ……………. 49
Figure 6.5: Average Http delay ………………………………………………… 50

7
List of Tables

Table 2.1: Metrics list ………………………………………………………….. 16

Table 3.1: Network simulators list ………………………………………......…. 21
Table 3.2: Application schedule ……………………………………………...... 29
Table 3.3: CBR parameters used for simulation ……………………………...... 29
Table 3.4: FTP parameters used for simulation ………………………………... 29
Table 3.5: HTTP parameters used for simulation …………………………….... 30
Table 3.6: Hardware hosting simulation ……………………………………….. 31
Table 3.7: Software used in the simulation …………………………………….. 32
Table 4.1: Average CBR throughput …………………………………………... 33
Table 4.2: Average CBR delay ……………………………………………….... 34
Table 4.3: T-test for CBR throughput, packet size 1024 …................................. 35
Table 4.4: T test for CBR delay, packet size 512……………………………….. 37
Table 4.5: T test for CBR delay, packet size 1024 and 2048…………………... 37
Table 4.6: Comments and explanation for 1024 and 2048………….………….. 38
Table 5.1: Average FTP throughput ………………………………………….... 39
Table 5.2: Average FTP delay …………………………………………………. 40
Table 5.3: T-test for FTP throughput packet size 512 …………………………. 41
Table 5.4: T-test for TCP throughput packet size 1024 and 2048………….…... 42
Table 5.5: Finding’s comments for Window size 10, packet size 1024 and 2048
…………………………………………………………………………………... 42
Table 5.6: Finding’s comments for Window size 50………………………...…. 43
Table 5.7: Finding’s comments for Window size 100 …………….………….... 44
Table 6.1: Average HTTP throughput …………………………………………. 46
Table 6.2: Average HTTP delay …………………………………………….…..46
Table 6.3: HTTP’s finding comments …………………………..………….….. 48
Table 6.4: T-test for HTTP delay, connection number 5 …………..…….……. 50

8
List of abbreviations and acronyms

CBR : Constant Bit Rate

FTP : File Transfer Protocol

HTTP : Hypertext Transfer Protocol

IP : Internet Protocol

ISO : International Organization for Standardization

J-SIM : Java Sim

LAN : Local Area Network

MAN : Metropolitan Area Network

MPLS : Multiprotocol Label Switching

NS2 : Network Simulator version 2

OSI : Open System Interconnection

OTCL : Object Oriented Tool Command Line

QoS : Quality of Service

TCL : Tool Command Language

TCP : Transport Control Protocol

UDP : Data Protocol

VPN : Virtual Private Network

WAN : Wide Area Network

WLAN : Wireless Local Area Network

9
Chapter 1: Introduction

1.1 Background and motivation

A virtual private network (VPN) is a network device or series of devices used to
interconnect various geographically separated sites, or connect remote users back
to a home network. As VPNs continue to evolve with a growing number of
options, they become more attractive to small business owners in terms of
providing secured communication over a public infrastructure like the internet
(Narayan et al., 2009). VPNs also offer benefits such as flexibility, connectivity
and security at a cheap cost. Enterprises benefit from VPN in reducing the cost,
increasing the scalability, and increasing the productivity without compromising
the security (Fisli, 2005).

The aim of this present research is to investigate the performance of networks

using Network Simulator version 2 (NS2), a network simulation environment.
Two setups are considered where in the first instance the network is without VPN,
and in the second the network is coupled with a VPN. Per scenario, various
trigging parameters are changed alternatively in the simulation. The computed and
calculated throughput and delay of some common network applications like
Constant Bit Rate (CBR), File Transfer Protocol (FTP), and Hypertext Transfer
Protocol (HTTP) are assessed using statistical analysis. The assessment’s results
of these two scenarios are compared to interpret the impact.

1.2 Problem statement

In the current era of information, VPN plays a crucial role in the sense that
productivity is not compromised as users can get connected from anywhere the
internet. However, it has been noticed practically that communication with and
without VPN are different with the former being slow. This weakness has brought
about the need to establish the impact of a VPN on the network performance.

10
1.3 Research objectives
From the previous section, it is understood that in terms of performance, a VPN-
based communication system is relatively slow. It is the purpose of this research
to determine the impact and extent of VPN usage on network performance. The
specific objectives are:

 To identify common network performance metrics and evaluate its

VPN’s impact
 To compare the VPN’s impact from different common network
applications perspective
 To recommend on how to reduce this VPN network performance
impact

1.4 Research Questions

Based on the above research objectives and considering that the most common
network performance metrics are throughput and delay (Hasib, 2006), the work
seeks to answer the following questions:

 What is the common network performance metrics, do VPNs impact

them?
 Does VPN’s impact vary differently with a different type of
applications or protocols?
 What needs to be done in order to reduce the identified impact?

1.5 Scope of the study

The research study is simulation-based; as such, no real life testing or
implementation is done. It has not been possible to run a test bed experiment on a
specific network due to logistical limitations.
The simulations are set to cover three applications: HTTP, FTP and CBR. In
terms of traffic settings parameters, packet size, window size, transfer rate, and a

11
number of connections are used. And for all simulations, the default bandwidth of
100 Mbps is used to mimic as closely as possible to real-world networks.

1.6 Layout of the dissertation

This dissertation comprises seven chapters. Chapter one presents the background
to the study, problem statement, objectives, research questions followed by the
scope of the research, and the structure of the dissertation.

In Chapter two, a review of relevant studies to the topic is done. The chapter
introduces the importance of network layer structures. Topics relating to VPNs,
network simulation, and network performances are covered.

Chapter three describes in detail the research methodology, procedures and

techniques used in this dissertation. Chapters four through six deals respectively
with the statistical analysis and interpretation of the CBR, FTP and HTTP
simulation data gathered.

Finally, chapter seven conclude the dissertation by summarising the relevant

findings. Future work is also identified and recommendations are made in line
with this.

12
Chapter 2: Literature review

2.1 Introduction
This chapter gives an overview of the research and the prior research studies that
have been done regarding the research stream. It considers what questions have
been addressed and answered before, and what results have been produced. For
a strong foundation and basis for this research, some definitions and the network
reference model networking concept are presented first. This is followed by the
main component of the research, network performance. Furthermore, to identify
the research gaps in order to develop the research study, an overview of the
current trends in research and industry related to the VPN and its impacts is
discussed. Literature was collected from credible resources like journals, research
reports, books and internet searches. The review of literature starts with the next
section which is the definition of concepts.

2.2 Clarification of concepts

Since the research focus is mainly on the term VPN, it is fair enough that research
starts by defining this term. VPN has come to mean many things to many people.
There are thousands of books that define it in whichever way serves the context.
For simplicity of this research, the VPN is defined as a network device or series of
devices used to interconnect various geographically separated sites, or connect
remote users back to a home network (Tan, 2003). To accomplish this
functionality, the VPN server uses two network interface cards. One connected to
the intranet network and another one connected to the public network. Staff
connected to the internet, find a VPN server, and then connect to the corporate
intranet. VPN actually uses a common link on the internet and it is essentially the
use of encryption technology to encapsulate a data communication tunnel which
transmits data on the public internet. The internet has a characteristic that it is
structured with a uniformed network stack (TCP/IP) that all the different layers
technologies can be implemented differently but with a uniformed interface with

13
their neighboured layers, Open System Interconnection Reference Model. The
next section explains the reference model.

2.3 Open system interconnection

At the beginning of the internet, there were many communication frameworks.
The result was if one’s framework is chosen, then the corresponding product
should be the only one with the same framework for network connection to
happen. Fortunately, the International Organization for Standardization (ISO)
published the Open System Interconnection reference model (OSI). It is a layered
structure aimed to provide a framework of open system protocol.

Figure 2.1 Basic architecture of OSI vs. TCP/IP

Source taken from Boston Technology Campus (2016). CompTIA Network+
Certification (Exam N10-006) official study guide: Study Notes

Using a layered structure as shown in Figure 2.1, the functionalities of a computer

network is organised in a stack of layers. This is to facilitate design, flexible
implementation, and intercommunication of computer network.
There is a virtual link between the corresponding layers in two communicating
nodes. Data flow occurs in a vertical fashion: from the highest layer to the lowest

14
layer in a node, and then through the physical link to reach the lowest layer at the
other node, and then following upwards to reach the highest layer in the stack.
Each layer represents a well-defined and specific part of the system and provides
certain services to the above layer. Accessible (by the upper layers) through
interfaces, these services usually define what should be done in terms of network
operations, but does not specifically define how such things are implemented.
The network simulation is no exception to the rule: it has to be able to incorporate
the features of the OSI reference model. It also caters for future packages to be
included and run transparently without disturbing existing components.
Based on the user demand at the application layer, a sending agent at the transport
layer constructs packets and transmits them to a receiving agent at the
corresponding transport layer through a low-level network.

2.3.1 Application layer

An application sits on top of the transport layer. It informs the attached agent (in
transport layer) of user demand. The simulation environment used in this research
classifies applications into traffic generators and simulated applications. A traffic
generator creates user demand at a specific time as scheduled, while a simulated
application acts all the way as if the application is running (Issariyakul and
Hossain, 2008).
To make the research design as close as possible to the real world network, not
only the CBR and FTP is used, but PackMimeHTTP as well. PackMineHTTP is
an NS2 object that drives the generation of HTTP traffic. Each PackMimeHTTP
object controls the operation of two types of applications, a PackMimeHTTP
server application and a PackMimeHTTP client application. Each of these
applications is connected to a TCP agent.

2.3.2 Transport layer

According to Figure 2.1, the network layer is responsible for delivering packets to
any destination node within the network, and consequently, the routing of packets
is one of the tasks of its protocols. IP is used by the two main protocols of the

15
transport layer, UDP and TCP, both providing packet flows between two hosts for
application layer protocols.
UDP and TCP are among the most widely used transport layer protocols. There is
a significant difference between these two protocols. While TCP provides a
reliable transmission, the much simpler UDP only transmits packets and does not
guarantee their delivery. In NS2, UDP is used for example by an application like
CBR, while TCP is used by FTP and HTTP.

2.4 Network performance

The network can be defined as a collection of devices that inter-communicate,
transmit, and receive data amongst themselves. An informal functional approach
to the definition of network performance is “the speed of traffic through the
network”. Is speed everything? The answer is that speed is not everything. Also,
the ability of a network to support large volumes of data transfer in the network,
for instance, is part of the picture and as well some external clock source. Still,
this is not yet everything for network performance. Network performance can be
referred as a measure of service quality; it has been tested by different parameters.
These parameters are called performance metrics and can be measured to evaluate
the performance of a network. Table 2.1 displays a list of some metrics.

Table 2.1: Metrics list

Quantitative metrics Qualitative metrics Other aspects

Bandwidth Good sound CPU usage
Throughput Bad picture Memory usage
Latency Slow response … Chassis temperature
Jitter Packet reordering …
Packet loss …
Hasib (2006) argued that the most common network performance metrics are
throughput and delay. Throughput (expressed in Bytes/second) can be defined as a
measure of the data transfer per unit time between a sender and a receiver. Delay

16
(expressed in millisecond), on the other hand, is the time needed for a packet to
travel.

These two metrics are the only considered network performance metrics of this
research.

2.5 Review of previous researches done on VPN impact

Previous studies have been done to evaluate the performance and impact of VPN
systems. The literature review reveals two aspects of the study that can be
categorised in two, the impacting aspect and the impacted aspect.

2.5.1 The impacting aspect review

The focus of this review’s aspect is on VPN factors affecting the performance of a
network. The main factors of this aspect are: security services, algorithms,
hardware, and software, just to name a few.
Extensive studies have been done on account of the impacting aspect.
Narayan et al. (2009) discussed the choice of VPN technologies and the
associated algorithm that affects network performance. They proved that VPN
protocols give different network performance metrics values for different
combinations of operating systems, protocols and algorithms.
Kumar et al. (2002) presented a scenario consisting of various algorithms for
provisioning VPNs. They concluded that even for the simple scenario in which
network links are assumed to have infinite capacity, the VPN’s algorithm’s
computation does have an impact on performance. They then proposed an
improved model, but degradation still persisted.
Passito et al. (2005) carried out an analysis of IPSec overhead in 802.1 lb
networks which showed a high overload in TCP and UDP traffic. The higher the
security level, the higher the system overhead.
Lyu and Lau (2000) created a real-world environment and set up a secured
firewall system. The objective was to study the performance of the firewall

17
system configured with different security policies and controls. Tests were
designed for security verification and performance measurement.
Based on the test results; it was concluded that the firewall technology is a major
factor for performance degradation. To put it another way, if frequent connections
with small size data are required in communication, enhanced firewall security is
likely to bring about performance degradation to the private network. However,
their work disproved the existence of a security-performance relationship. The
same level of security at firewall was found to be achievable with different
firewall technologies. Inconsistent levels of network performance were noted as a
result.
Pruthi (2009) presented a project and implementation of a WLAN with IPSec, an
overview of security issues. He constructed an IPSec based VPN between the
mobile node and the IPSec Gateway and investigated the performance of the UDP
and TCP protocols. It was found that variation in packet loss rate and network
throughput was affected.
Ashraf et al. (2009) were able to argue that even an unsecured VPN affects
network performance; hence, the QoS experienced by the end-to-end users.

2.5.2 The impacted aspect review

On the other hand, this review’s aspect focuses on the affected elements.
These elements are performance metrics such as bandwidth, throughput, packet
loss, delay and so on.
Speaking about the impacted aspect, Powell (2010) showed that the use of a VPN
for database transactions leads to a 446% delay of the query. Delays are also
experienced in e-mail and FTP transactions.
Agarwal and Wang (2005) were able to demonstrate that throughput and delay are
affected by different policies. They showed that authentication time may be a
factor contributing towards QoS degradation than encryption overhead. This
ascribed to hardware becoming faster.
Bourdoucen et al. (2009) simulated wireless LAN for IEEE802.11g protocol. The
research group implemented a VPN by integrating an IPsec Virtual Private

18
Network technology. This was to secure the flow of traffic between the client and
the server; OPNET WLAN utility was used for the purpose.
Their study underscored the importance of Response Time and Load of the
Virtual Private Network over wireless LAN on performance.
Malik and Verma (2012) carried out simulation in the OPNET environment. In
the simulation setup, voice and video applications were employed under two
scenarios: without VPN and with VPN. Data relating to packet delay, traffic
received and traffic sent was collected. The two researchers were able to
demonstrate that the use of VPN increases the level of security; and in contrast,
decreases network performance.
Rahimi et al. (2009) used J-Sim to simulate an MPLS for different packet drops,
delays and throughputs. The study demonstrated a decrease in throughput and
proportional losses in terms of packet losses.

2.6 Summarised findings

It is demonstrated that several studies have been done on the impact of VPN on
network performance. The literature review performed concurs that performance
is amongst the most important features in a network, regardless of the
architecture. And different network architectures exist; LAN (Local Area
Network), Metropolitan Area Network (MAN), Wide Area Network (WAN),
VPN, Cloud, just to name a few. From all these architectures, VPN appears to
take the lead.
Research on network performance in a VPN context has been previously
considered. The key finding was that in most of the cases, there was a sensible
variation over delay and network throughput. Throughput at the destination host
was lower than that at the traffic source. And in terms of delays, traffic
experienced proportional delays.
Besides, the VPN network security was a function of its algorithm.
The literature review also showed little research on the impacted factor (network
performance). These factors may grow in importance in future. With powerful

19
simulations like NS2, this type of study is becoming possible. It is the intention of
this research project to explore this and pave the way for further work.

20
Chapter 3: Research methodology, procedures and
techniques

An account of a review of relevant literature was given in chapter 2 and

subsequently, this chapter is devoted to a discussion of the design of the empirical
investigation. Aspects that are discussed include the methodology of the research,
the research instrument used, the data collection, the challenges and difficulties
encountered. This chapter gives an overview of the research methodology
adopted, the data collection technique and the simulation programme used.

3.1 Network simulations

Simulation can be defined as the recreation of real world scenarios using
computer programmes. The technique enables one to perform extensive
experimentation while relying on mathematical models of the system under
investigation. The major advantage of simulations is that features of the model
can be changed and the corresponding results are analysed cost-effectively in
comparison to the real life experiments.
Network simulation tools can be categorised based on whether they are
commercial or free. In the first instance, the network simulators do not provide the
source code to the public for free; the users have to pay for the license. In the
second instance, the open source network simulator has the advantage that the
source code is open and free. The source code is also flexible and develops faster
from the contribution of everyone involved. There is a wide variety of network
simulators; Table 3.1 provides a list of the most common network simulators in
the market.

21
 Table 3.1: Network simulators list

Commercial Open source

Matlab:(Physiomeproject,2015) NS2: (Issariyakul and Hossain, 2008)

Opnet: (Riverbed, 2012) Omnet++: (Omnet++, 2015)
J-Sim: (Physiome project, 2015)

NS2 is arguably the most popular in academia because it is open-source, it has

plenty of network and protocol objects library, and it is targeted at network
research. From all the above, the research chooses NS2 as the research simulation.
NS2 is a discrete event simulator where the advance of time depends on the
timing of events which are maintained by a scheduler. NS2 is a combination of
C++ language and OTCL script. The OTCL script is an object-oriented tool
command line interpreter which interprets the scripts line by line the code written
in notepad or in Microsoft-word etc. and save it by .TCL extension. Tool
Command Language (TCL) is a language with a very simple syntax and allows a
very easy integration with other languages. NS2 is installed on Linux and it has a
command line interface.
Figure 3.1 shows the NS2 command line interface in Ubuntu Linux.

Figure 3.1: NS2 command line interface

22
The Linux command “Vim” (Figure 3.2) is used to create or open a file script in
NS2. This can be achieved alternatively by opening a notepad file and save the
file under the NS folder using a .tcl extension.

Figure 3.2: Vim command to create or open tcl script

NS2 simulation starts with the command “set ns [new Simulator]” which is the
first line in the TCL script to initialize the simulation. From initialisation, we
define nodes, links, queue, topology, agents, and applications and trace files using
varieties of command lines. The simulation can then begin using the command
“$ns run”.
Figure 3.3 shows some of the command lines as it is inserted in a TCL script file.

Figure 3.3: Sample commands for TCL script

The NS2 output is either text-based or animation-based data. In order to analyse

this data and study the network behaviour, the text-based data is formatted into a
convenient form.

23
3.2 Simulation Setup
Considering that various factors contribute to network performance degradation,
the test bed used a wired connection as opposed to the wireless one. Hence, some
obvious initial degradation is cut off and makes the environment ideal. The
simulation is deployed in two scenarios. Figure 3.4 presents the first scenario
where the connection between the two nodes is a direct line, no VPN involved.
This scenario consists of just 2 normal interconnected nodes, the research’s
baseline.

Figure 3.4: Scenario 1: Network without VPN

The second scenario (Figure 3.5), the VPN comes in the picture. This scenario
consists of four interconnected nodes (2 hosts plus 2 extra nodes: routers with
routing activated which form a VPN).

Figure 3.5: Scenario 2: Network with VPN

24
Figure 3.6 presents all processes and steps followed by the research from the
implementation all the way to the findings. The writing of the simulation script,
the running of the script, the data collection and its interpretation

Figure 3.6: research flowchart

The simulation starts with the creation of the two scenarios, we define and setup
scenario without VPN and scenario with VPN in within NS2. See index

25
Both scenarios are run using alternatively the two protocols which are TCP and
UDP. Simulation’s data are collected by protocol and by scenario using trace files.
The collected data are then analysed and compared accordingly. If any
significance of the difference is observed, then the impact existence is concluded.
Otherwise no impact
To setup the test bed simulation, a CBR, FTP and HTTP applications are
respectively generated following procedures in Figures 3.7 – 3.12.

3.3 Simulation Procedures

3.3.1 Procedure case scenario 1

For CBR application in Figure 3.7, the following procedures are implemented:

 Create node1 and node4:

set node1 [ns node], set node4 [ns node]
 Create link between node1 and node4:
$ns duplex-link $node1 $node4 100Mb 10ms DropTail
 Attach UDP and Sink to nodes
$ns attach-agent $node1 $udp
$ns attach-agent $node4 $lostpktsink_udp
 Attach CBR application to UDP agent
$cbr attach-agent $udp
 Connect agent to sink
$ns connect $udp $lostpktsink_udp

Figure 3.7: CBR application over UDP in scenario without VPN

26
In Figure 3.8, FTP application uses the same procedure as the previous.
The difference is the change in protocol and application itself; TCP and FTP as
opposed respectively to the former, UDP and CBR.

Figure 3.8: FTP application over TCP in scenario without VPN

In the HTTP implementation (Figure 3.9), beside the creation of nodes and link,
the following procedure is applied:
 Create Client
$pm set-client $node1
 Create Server
$pm set-server $node4
 Create traffic
$pm set-http-1.1

Figure 3.9: PackMimeHTTP Architecture in scenario without VPN

27
3.3.2 Procedure case scenario 2
Figure 3.10, same as scenario 1 above, plus two extra nodes in between. As well
activate routing as follow:

 Create 2 extra nodes (node2 and node3)

 Create link between nodes
$ns duplex-link $node1 $node2 100Mb 10ms DropTail
$ns duplex-link $node2 $node3 100Mb 10ms DropTail
$ns duplex-link $node3 $node4 100Mb 10ms DropTail
 Activate routing
$ns rtproto DV

Figure 3.10: CBR application over UDP in scenario with VPN

The FTP uses same procedure than the above CBR except the change in protocol
and application; TCP and FTP respectively (see Figure 3.11).

Figure 3.11: FTP application over TCP in scenario with VPN

28
Figure 3.12 HTTP, is same as scenario 1 HTTP except the two extra nodes and
routing.

Figure 3.12: PackMimeHTTP Architecture in scenario with VPN

3.4 Simulation schedule and parameters

In this section, the scheduling of events is emphasised. Start and stop time of the
simulation. As well as the parameters which trigger the results are introduced,
packet size, window size, transfers rate and number of connections.

3.4.1 Schedule
The initializing command “set ns [new simulator]” creates an event scheduler, and
events are then scheduled using the format:
$ns at <time> <event>
The scheduler starts through the command:
$ns run
In this simulation, the beginning and end of the three applications are set as
follow:
$ns at 2 "$cbr start"
$ns at 7 "$cbr stop"
$ns at 8 "$ftp start"
$ns at 13 "$ftp stop"
$ns at 14 "$pm start"
$ns at 19 "$pm stop"
$ns at 20 "finish"
$ns run

29
Table 3.2 gives schedule resume for the three applications.
Table 3.2: Application schedule
Application Start Stop Simulation
Time Time
CBR 2s 7s Models a user demand to periodically transmit
data. Example: video and VoIP streams
FTP 8s 13 s Models a bulk data transfer
HTTP 14 s 19 s Models a traffic generation on a link shared by
many web clients and servers

3.4.2 Simulation parameters

For each scenario we ran the applications one by one, changing gradually the
combination of settings (Tables 3.3, Tables 3.4 and Table 3.5)

Table 3.3: CBR parameters used for simulation

Packet Size (Bytes)
512 1024 2048
CBR Transfer
rate (Mb)
10 X X X
100 X X X
1000 X X X

Table 3.4: FTP parameters used for simulation

Packet Size (Bytes)
512 1024 2048
TCP window size
(Mb)
10 X X X
50 X X X
100 X X X

30
 Table 3.5: HTTP parameters used for simulation

HTTP number of connection

5
10
15

3.5 Data collection

In this section, the recording and data collection method of the research is
explained.
A Tcl script is compiled out of various NS2commands (See Appendices A, B, and
C). This Tcl script is run, the trace result collected and stored in a text format file.
Furthermore, the text format’s trace is imported, processed and organised in
Excel® format. The Excel® format is ready to be used for graphs and statistics
such as means (averages). Trace file examples in an original format are illustrated
in Figures 3.13 and 3.14

Figure 3.13: Default NS2 trace file

Figure 3.14: Customised TCP throughput trace file

31
3.6 Statistical procedures and techniques
Data collected were analysed using a hypothesis t-test. A hypothesis test is a
process by which an analyst tests a statistical hypothesis. The hypothesis to be
tested is called the null hypothesis. The null hypothesis is rejected if the value of
the test statistic falls within the critical region and is not rejected otherwise. The
methodology used by the analyst depends on the nature of the data used, and the
goals of the analysis. The goal as stipulated above is to either accept or reject the
null hypothesis which is in our case the “no impact”. In order to accept or reject
this hypothesis, the research uses a t-test which is one of a number of hypothesis
tests to depict statistical differences between two population means. Assuming
that the variances are unequal, the research applies a two-tail test (inequality) of
the t-test which states that:
If “(t Stat) < - (t Critical two-tail)”Or “(t Stat) > (t Critical two-tail)”, we reject the
null hypothesis. Otherwise; “- (t Critical two-tail) < (t Stat) < (t Critical two-tail)”,
the null hypothesis is accepted
To be noted:
 “(t Stat)” : t value calculated from the data
 “(t Critical two-tail)”: the tabulated value; the value that we would need to
exceed in order for the difference between the means to be significant at
the 5% level (Ross, 2010; Neil, 2012)

3.7 Hardware specifications

A laptop of specifications as shown in Table 3.6 was used as simulation host.

Table 3.6: Hardware hosting simulation

Brand Hewlett Packard
Model Hp ProBook 4530
Processor name Intel Core i3
Processor speeds 2.20 GHz
Ram 4.00 GB
Storage 500 GB
Ethernet Realtek PCIe

32
3.8 Software specifications
Four software were used in this study as shown in Table 3.7

Table 3.7: Software used in the simulation

Software used Role played

Microsoft Windows 7 Professional 32- Host operating system
bit
Oracle virtual box-4.3.12-93733-Win Virtual machine running on top of
windows
Ubuntu-14.04-server-i386 Linux version installed on the virtual
machine
NS2.35 Simulator Simulation software installed on
Ubuntu

3.9 Challenges and difficulties encountered

The no access to a real life experimental system was the main limitation of the

study. Besides, the use of a VM (Virtual Machine) created bugs which made the

Nam (Network animator) unable to visualise real world live traffic.

33
Chapter 4: Effects of VPN on CBR application

4.1 Introduction
This chapter presents an analysis and evaluation of VPN’s impact on CBR
application. The research benchmarks Scenario 1 as the reference point that later
on is referred back to compare against Scenario 2. This evaluation involves
comparing the means of various active measurements (Calyam et al., 2005) for
throughput and delay respectively. Results are presented and analysed using
graphs and hypothesis testing respectively

4.2 Collected results for CBR

Various tests are alternatively run using different aspects of transfer rates and
packet size as per table 3.3 above. With CBR sitting on UDP protocol, the
simulation results are summarised and presented in table 4.1 and table 4.2 below.

Table 4.1: Average CBR throughput

Packet size
512 Bytes 1024 Bytes 2048 Bytes
Transfer A B A B A B
rate 10 0.0073 0.0073 0.0145 0.0145 0.0291 0.0291
Transfer A B A B A B
rate 100 0.0073 0.0073 0.0145 0.0145 0.0291 0.0291
Transfer A B A B A B
rate 1000 0.0073 0.0073 0.0145 0.0145 0.0291 0.0291

34
 Table 4.2: Average CBR delay
Packet size
512 Bytes 1024 Bytes 2048 Bytes
Transfer A B A B A B
rate 10
0.01004 0.01671 0.01008 0.02012 0.01013 0.02218
Transfer A B A B A B
rate 100
0.01004 0.01671 0.01008 0.02012 0.01013 0.02218
Transfer A B A B A B
rate 1000
0.01004 0.01671 0.01008 0.02012 0.01013 0.02218

4.4 Performance evaluation

This section evaluates the results obtained from the research simulation in section
4.3. Gathered CBR’s throughput and delay data from the two different scenarios
are presented and analysed.

4.4.1 Throughput
One of the metrics analysed in this section is throughput, it uses two parameters
which are packet size and transfer rate.
It can be seen from Table 4.1 that column A equals column B for each packet size.
And as the three transfer rates provide same insight, a random transfer rate and
packet size is selected and its results presented on behalf of all. Figure 4.1 gives
us an average CBR throughput graph of three different packet sizes: 512, 1024
and 2048 Bytes.

35
 0.03500

Average throughput (Mbits/s)

0.03000

0.02500

0.02000

0.01500 A: Without VPN

0.01000 B: With VPN

0.00500

0.00000
512 Bytes 1024 Bytes 2048 Bytes
Packet size (Bytes)

Figure 4.1: Average CBR throughput

From figure 4.1 it is clear that column A equals column B. No change, thus
impact. Choosing randomly 1024 bytes’ data from appendix E, the t-test gives the
result in table 4.3

Table 4.3: Student’s t-test for CBR throughput, packet size 1024
t-test: two-sample assuming unequal variances
With VPN Without VPN
Mean 0.014545455 0.014545455
Variance 2.32727E-05 2.32727E-05
Observations 11 11
Hypothesized Mean 0
Difference
tSat 0
t Critical two-tail 2.085963447

From the above t-test table and applying the theory in section 3.6: “Statistical
procedures and techniques”, it is found that the condition
“- (t Critical two-tail) < (t Stat) < (t Critical two-tail)” is observed:

36
- 2.085963447 < 0 < 2.085963447. The null hypothesis is not rejected, no impact
on this throughput.

4.4.2 Delay
The delay has a different approach than throughput. Column A differs from B in
all the packet size cases, but the transfer rates present same insight for all 10, 100
and 1000 Mb. As observations for transfer rate provide the same insights, a
random transfer rate of 1000 MB is selected and results presented. Figure 4.4
shows the average CBR delay graph of transfer rate 1000 Mb, packet size 512,
1024 and 2048 Bytes.

0.02500
Average delay (Milliseconds)

0.02000

0.01500

A: Without VPN
0.01000
B: With VPN

0.00500

0.00000
512 Bytes 1024 Bytes 2048 Bytes
Packet size (Bytrs)

Figure 4.2: Average CBR delay of transfer rate 1000

From figure 4.2 above and among its three packet sizes, the 512 bytes is randomly
selected and a t-test is computed out of it using its data from appendix G to give
table 4.4

37
 Table 4.4: Student’s t-test for CBR delay, packet size 512
t-test: two-sample assuming unequal variances
With VPN Without VPN
Mean 0.016707967 0.010041
Variance 9.30844E-05 3.34363E-36
Observations 30 10
Hypothesized Mean 0
Difference
tSat -3.784866137
t Critical two-tail 2.045229642

Using the results from the above table in conjunction with the theory in section
3.6: “Statistical procedures and techniques”, it is observed that
“(t Stat) < - (t Critical two-tail)” of the theory is proven;
-3.784866137 < -2.045229642; the null hypothesis is rejected.
These two scenarios differ, thus existence of impact on the delay.
Packet size 1024 bytes and 2048 bytes follow the same pattern than 512 bytes
discussed above. The student’s t-test in Table 4.5 combines both two results

Table 4.5: Student’s t-test for CBR delay, packet size 1024 And 2048
t-test: two-sample assuming unequal variances
Packet size 1024 Packet size 2048
With Without With VPN Without VPN
VPN VPN
Mean 0.02012 0.01008 0.02218 0.01013
Variance 0.00011 1.05E-12 0.0001 1.55E-09
Observations 40 20 50 30
Hypothesized 0 0
Mean Difference
tSat -6.19529 -8.47857
t Critical two-tail 2.02269 2.00958

38
The results in Table 4.5 Show that the null hypothesis can be rejected. In both
cases the following condition “(t Stat) < - (t Critical two-tail)” is observed:
See comments in table 4.6

Table 4.6: Comments and explanation

Packet size 1024 Packet size 2048

(t Stat) < - (t Critical two-tail). “(t Stat) < - (t Critical two-tail)”
-6.19529 < -2.02269 -8.47857< -2.00958
The null hypothesis is rejected The null hypothesis is rejected
Thus impact Thus impact

In both Cases, the delay is affected

4.5 Summarised findings

Concerning throughput, all traffic injected was received regardless of packet size
or window size. It is proved that there is nothing lost in the UDP’s network
journey; consequently, throughput A and throughput B remain constant. No
impact observed for throughput.

On the delay account, the student’s test proved that the average delay in A
scenarios was not the same compared to the B scenarios. Thus, the delay was
influenced

39
Chapter 5: Effect of VPN on FTP application

5.1 Introduction
In this chapter, results of the investigation in line with the FTP application is
analysed, interpreted and presented. The same statistical tool, t-test of section 3.6
“Statistical procedures and techniques” is used to test the no impact hypothesis.

5.2 Simulation results for FTP

Tests are run using different aspects of window sizes and packet sizes as shown in
Table 3.4. With FTP using TCP protocol, the experimental results are summarised
and presented in Table 5.1 and Table 5.2 below.

Table 5.1: Average FTP throughput

Packet size
512 Bytes 1024 Bytes 2048 Bytes
Window A B A B A B
size 10 1.7857 0.6464 3.8164 1.2427 0.0291 0.0291
Window A B A B A B
size 50 9.7146 3.1382 18.9 6.0546 36.9309 11.8773
Window A B A B A B
size 100 19.5318 6.2052 37.6418 11.9573 73.4173 23.31

40
 Table 5.2: Average FTP delay
Packet size
512 Bytes 1024 Bytes 2048 Bytes
Window A B A B A B
size 10 0.01 0.03 0.01 0.03 0.01 0.03
Window A B A B A B
size 50 0.01 0.03 0.01 0.03 0.01 0.03
Window A B A B A B
size 100 0.01 0.03 0.01 0.03 0.01 0.03

5.4 Throughput’s performance evaluation

Window size is used to trigger results in turns with 10Kb, 50Kb and 100Kb

5.4.1 Window size 10 Kb

Figure 5.1 shows the average TCP throughput graph of window size 10 Kb,
packet size 512, 1024 and 2048 Bytes.

25
Average throughput (Mbits/s)

20

15

A: Without VPN
10
B: With VPN

0
512 Bytes 1024 Bytes 2048 Bytes
Packet size (bytes)

Figure 5.1: Average TCP throughput of window size 10

41
To conclude this case, the t-test is used for each of the three different packet sizes.
For these different situations, table 5.3 and table 5.4 are respectively presented as
follow:

Table 5.3: t-test for FTP throughput, packet size 512

t-test: two-sample assuming unequal variances
With VPN Without VPN
Mean 0.646363636 1.785727273
Variance 0.053165455 0.746839218
Observations 11 11
Hypothesised Mean 0
Difference
tSat 4.2248611
t Critical two-tail 2.20098516

Referring to the statistical tool of section 3.6, the below condition of hypothesis
rejection is met:
“(t Stat) > (t Critical two-tail)” which from the table gives
4.2248611 > 2.20098516
This implies the existence of a difference in Scenarios A and B. There is an
impact on this throughput case.
Follows is the window size test of still 10 Kb but the packet size is changed from
512 to 1024 and 2048 Bytes. The results of a combined t-test are shown in Table
5.4.

42
 Table 5.4: T-test for FTP throughput, packet size 1024 and 2048
t-test: two-sample assuming unequal variances
Packet size 1024 Packet size 2048
With VPN Without With VPN Without
VPN VPN
Mean 1.242727273 3.816363636 7.4518182 2.4409091
Variance 0.197861818 1.621405455 6.1771164 0.7655291
Observations 11 11 11 11
Hypothesized 0 0
Mean Difference
tSat 6.328419693 6.3074
t Critical two-tail 2.17881283 2.1788128

From table 5.4 the condition of “(t Stat) > (t Critical two-tail)” is met for both
1024 bytes and 2048 bytes as displayed in table 5.5

Table 5.5: Finding’s comments for Window size 10, packet size 1024 and
2048

Packet size 1024 Packet size 2048

(t Stat) > (t Critical two-tail) “(t Stat) > (t Critical two-tail)”
6.328419693 > 2.17881283 6.3074 > 2.1788128
The null hypothesis is rejected The null hypothesis is rejected
Thus impact Thus impact

The null hypothesis does not stand. The observation done is that the two scenarios
average throughputs are different in both cases. This indicates that throughput is
affected. This is enough evidence to agree on the inequality of the average
throughput between scenarios. There is an impact on these throughput cases.

43
5.4.2 Window size 50
Below (Figure 5.2) is a pictorial comparison between the average scenario TCP
throughput of window size 50 Kb, packet size 512, 1024 and 2048 Bytes.

40
35
Average throughput (Mbits/s)

30
25
20
A: Without VPN
15
B: With VPN
10
5
0
512 Bytes 1024 Bytes 2048 Bytes
Packet size (bytes)

Figure 5.2: Average TCP throughput of window size 50

Not to be repetitive, we present straight the findings. These findings verify the “ (t
Stat) > (t Critical two-tail)” condition as proven in table 5.6

Table 5.6: Finding’s comments for Window size 50

Packet size 512 “(t Stat) > (t Critical two-tail)”
6.200745254 > 2.160368656
The null hypothesis is rejected
Thus impact
Packet size 1024 “(t Stat) > (t Critical two-tail)”
6.24175476 > 2.160368656
The null hypothesis is rejected
Thus impact
Packet size 2048 “(t Stat) > (t Critical two-tail)”
6.2213516 > 2.1603687
The null hypothesis is rejected
Thus impact

44
5.4.3 Window size 100
For the next window size which is 100 Kb, we present it visual representation as
shown in figure 5.3 and give it summary interpretation as per the t-test null
hypothesis.

80
70
Average throughput (Mbits/s)

60
50
40
A: Without VPN
30
B: With VPN
20
10
0
512 Bytes 1024 Bytes 2048 Bytes
Packet size (bytes)

Figure 5.3: Average TCP throughput of window size 100

Examining this graph, it is found that the case follows the same pattern as the
previous two and the “(t Stat) > (t Critical two-tail)” condition is met for all the
three packet sizes

Table 5.7: Finding’s comments for Window size 100

Packet size 512 6.179818204 > 2.160368656
The null hypothesis is rejected.
Impact
Packet size 1024 6.177265158 > 2.160368656
The null hypothesis is rejected.
Impact
Packet size 2048 6.1577338 > 2.1603687
The null hypothesis is rejected.
Impact

45
5.5 Delay’s performance evaluation
Looking at data from figure 5.2, it is noticed that regardless of the window size
and packet size, we have the same pattern for all. Due to space limitation and not
to be exhaustive, we present and discuss the only one set of case on behalf of all
Figure 5.4 shows this average FTP delay graphically.

0.035

0.03
Average delay (Milliseconds)

0.025

0.02

0.015 A: Without VPN

B: With VPN
0.01

0.005

0
512 Bytes 1024 Bytes 2048 Bytes
Packet size (bytes)

Figure 5.4: Average FTP delay

In all the cases, it is observed that all the A columns are clearly smaller than the B
columns.

5.6 Summarised findings

From the above, this is enough evidence to imply on the inequality between the
two scenarios. Throughputs and delay are impacted in all cases.

46
Chapter 6: Effects of VPN on HTTP application

6.1 Introduction
This chapter analysis and evaluates the VPN’s impact on HTTP application using
PackMimeHTTP. Gathered results are presented and commented respectively
using graphs and hypothesis t-test as per section 3.6: “Statistical procedures and
techniques”.

6.2 Simulation results for HTTP

Simulations are run off different connection rates on the HTTP throughput and
delay of the network. Averages of the two scenarios are compared on the account
of throughput and delay. Connection rates of 5, 10, and 15 are in turns the triggers
parameters that are used in this simulation.
Gathered experimental results are summarised and presented in table 6.1 and 6.2.

Table 6.1: Average HTTP throughput

A: Without VPN B: With VPN

Connection rate 5 49.49358225 17.139651
Connection rate 10 39.46256811 16.583419
Connection rate 15 44.45390085 16.170693

Table 6.2: Average HTTP delay

A: Without VPN B: With VPN

Connection rate 5 0.01 0.03
Connection rate 10 0.01 0.03
Connection rate 15 0.01 0.03

47
6.4 Performance evaluation
This evaluation is presented for the account of throughput and delay. Our protocol
of interest in this specific case is the TCP which is the HTTP application’s
backbone.

6.4.1 Throughput
The average HTTP throughput graph of connection rate 5, 10 and 15 is presented
in figure 6.1.

60

50
HTTP average throusghput
(Mbits/second)

40

30
A: Without VPN
20 B: With VPN

10

0
#5 #10 #15
Number of connection rate

Figure 6.1: Average Http throughput

From the above graphical representation, it is found that the “(t Stat) > (t Critical
two-tail)” condition is met for all the three connection rate numbers as per the
table below.

48
 Table 6.3: HTTP finding comments
Connection rate 5 3.132110036 > 2.004044783
The null hypothesis is rejected.
Impact
Connection rate 10 3.955371321 > 1.980992298
The null hypothesis is rejected.
Impact
Connection rate 15 5.572167228 > 1.974715786
The null hypothesis is rejected.
Impact

To emphasise this difference between scenario A and scenario B, the instant

throughput versus time graph is introduced. When comparing these two scenarios
instant throughput, the shapes change their curves
Figure 6.2, figure 6.3, and figure 6.4 reflect these changes.

450

400
Instant trhroughput (Mbit/s)

350

300

250 No VPN
200 With VPN

150

100

50

Figure 6.2: Comparison of instant throughput, connection rate 5

49
 400

350

Instant throughput (Mbit/s)

300

250
No VPN
200
With VPN
150

100

50

Figure 6.3: Comparison of instant throughput, connection rate 10

400

350
Instant throughput (Mbit/s)

300

250
No VPN
200
With VPN
150

100

50

Figure 6.4: Comparison of instant throughput, connection rate 15

From the instant throughput perspective, it is observed that the effect of VPN
brings a decrease in the link utilisation and capacity. The results of this process
are viewable in the above figures and tables.
There is a difference, thus impact but no linear relationship exists between
scenario A and scenario B.

50
6.4.2 Delay
HTTP spends more time waiting than it does transfer data.
For this raison, the HTTP delay is an important metric to look after.
Figure 6.5 shows an average HTTP delay graph of connection rate 5, 10 and 15.

0.035

0.03
HTTP average delay

0.025
(Milliseconds)

0.02

0.015 A: Without VPN

B: With VPN
0.01

0.005

0
#5 #10 #15
Number of connection rate

Figure 6.5: Average Http delay

Table 6.4 gives the computed results which are the same for all the three number
of connections.

Table 6.4: T-test for HTTP delay, number of connection 5

t-test: two-sample assuming unequal variances
With VPN Without VPN
Mean 0.029840157 0.010050758
Variance 5.0398E-06 1.01529E-08
Observations 1350 1345
Hypothesized Mean 0
Difference
tSat -323.5595961
t Critical two-tail 1.961717572

51
The t-test condition meeting the requirement of the no rejection of the null
hypothesis as stipulated in section 3.6 is:

“(t Stat) < - (t Critical two-tail)”

-323.5595961 < -1.961717572; the null hypothesis is rejected.

These two scenarios differ significantly, thus the existence of impact on the delay.

6.5 Summarised findings

Both HTTP throughput and delay are impacted by the VPN, but these impacts at
the VPN level are not proportional to the number of connection.

52
Chapter 7: Conclusion and recommendations

7.1 Introduction
VPN plays an important role in the industry in the sense that it helps keep
productivity life. From where ever, one can always connect back to the office and
uses resources. However, it is understood that in terms of performance, a VPN-
based system is relatively slow when compares with a LAN direct connection.
This research commuted to determine the impact of VPN usage on network
performance.

7.2 Summary and conclusion

The goal of this research was to study network performance impact due to the use
of VPNs. To this end, simulations were set in the NS2 environment, a software
simulator for studying network systems. Relevant data output was generated and
collected under various testing conditions. Results were analysed and compared in
terms of throughput and delay averages using Student’s t-test. For the different
simulation conditions, a 100 Mbps was considered. In all the cases, the two
scenarios were simulated: network without VPN and network with VPN.

The main specific objectives were 1) to identify common network performance

metrics and evaluate its VPN’s impact, 2) compare the VPN’s impact from
different common network applications perspective and 3) to recommend on how
to reduce this VPN network performance impact.

Therefore, as to respond to the research questions the following conclusions can

be said:

Firstly, the research produced a number of findings that have helped to generate a
deeper understanding of how much the network performances are impacted by the
VPN.

For the first objective, the results revealed that the common network performance
metrics are throughput and delay. In terms of throughput, the UDP throughput

53
was found to be unaffected by the VPN regardless of the CBR transfer rates and
packet sizes tested. However, it was learned that the TCP throughput, as opposed
to the UDP, was influenced by the VPN. This influence was a function of both
packet size and window size.

In terms of the delay, it was learned that the UDP delay was sensitive to the VPN.
Packet size was one of the influential parameters for this impact. Moreover, it was
discovered that the transfer rate does not influence the UDP delay.

The TCP delay was found to be sensitive to the VPN and packet size was the only
one influential parameter for this sensitivity. Window size did not influence the
TCP delay.

As for the HTTP, it was learned that the HTTP throughput and HTTP delay was
affected by the VPN. The impact was related to the Connection rate but no linear
relationship exists between scenarios.

In achieving objective 2, we had a clear view and understanding that the impact of
the VPN depends on the protocol used. UDP is a datagram packet which has less
overhead. TCP, however, is a connection-based protocol which assumes an
unreliable medium. Consequently, the TCP has more overhead and encounters
adverse effects. Results proved in general that TCP throughput as opposed to
UDP throughput measured in scenario A is very much lower compared to the TCP
throughput measured in scenario B. Thus the FTP throughput and HTTP
throughput which are applications sitting on TCP protocol are impacted but the
CBR application sitting on UDP is not. Hence CBR, FTP and HTTP are
differently influenced.

For objective 3, the increase in delay is justified by the decrease in throughput and
vice versa. These two conditions can be accordingly rectified or improved by
choosing proper hardware, proper software, well - selected encryption algorithms
at the VPN level, and setting up configuration parameters of likes of small
window size, small packet size and small connection rate.

54
Although it can be improved by using faster hardware and better algorithms,
throughput is a critical performance metric which will limit the number of users
whom the VPN can support. Thus, higher throughput is better.

The more you decrease the packet size and window size, the less is the impact.

7.3 Recommendations and future work

The design should be extended so as to include other factors that are known to
potentially affect performance. For this, an extension of the simulation script is
required. The use of other simulation packages should be explored in addition to
NS2.

In the case of a simulation is used, this should be run on a physical machine rather
than a VM.

Finally, a validation programme whereby live experiments are carried out is

necessary to complete the work.

References
Agarwal, A.K., Wang, W. 2005‘Measuring performance impact of security
protocols in wireless local area networks’, 2nd International Conference on
Broadband Networks, 2005. North Carolina: IEEE

Ashraf, M.T. Davies, J.N Grout, V. 2009, An investigation into the effect of
security on performance in a VoIP network, Glyndwr University Research Online,
viewed 27 August 2014,
<www.glyndwr.ac.uk/computing/research/pubs/sein_adg.pdf >

Bourdoucen, H., Al Naamany, A., & Al Kalbani, A. 2009, ‘Impact of

implementing VPN to secure Wireless LAN’. International Journal of Computer,
Information and Systems Science, vol. 3, no. 1, pp. 17

55
Calyam, P., Krymskiy, D., Sridharan, M., & Schopis, P., 2005 ‘Active and passive
measurements on campus, regional and national network backbone paths’. In 14th
International Conference Computer, Communications and Networks. ICCCN
2005. USA: IEEE

Boston Technology Campus 2016, ComTIA Network+ Certification Official study

guide: Study Notes, g525eng ver075. Gtslearning

Fisli, R. 2005, ‘Secure corporate communications over VPN-based WANs’,

Master’s degree’, Royal Institute of Technology

Hasib, M. 2006, ‘Analysis of packet loss probing in packet networks’, PhD thesis,
University of London

Issariyakul, T. & Hossain, E. 2008, Introduction to network simulator NS2, 2nd

edition. Springer, New York

Kothari, C.R. 2004, Research methodology, methods and techniques, 2nd revised
edition, New Age International (Pty) Ltd, New Delhi

Kumar, A. Rastogi, R. & Silberschatz, A. 2002 ‘Algorithms for provisioning

virtual private petworks in the hose model’. In ACM Transactions on networking.
United States: IEEE

Lyu, M.R. & Lau, K.Y. 2000 ‘Firewall security: Policies, testing and
performance evaluation’. In the 24th Annual International Computer Software and
Applications Conference, COMPSAC 2000. Taipei: IEEE

Malik, A. & Verma, H. 2012, ‘Performance analysis of virtual private network for
securing voice and video traffic’, International Journal of Computer Applications,
vol. 46, no. 16, pp. 25 – 29

56
Mathworks 2015, Mathworks. Matlab academy. Viewed on 6 May 2015,
<http://www.mathworks.com>.

Narayan, S., Brooking, K., & de Vere, S. 2009, ‘Network performance analysis of
VPN protocols: An empirical comparison on different operating systems’. In
International Conference on Networks Security, Wireless Communications and
Trusted Computing, Auckland: IEEE

Neil, A.W. 2012, Introductory statistics, 9th Edition. Addison-Wesley, Arizona

State University: United States:

Omnet++ 2015, Discrete event simulator, Omnet. Viewed on 5 August, 2015

<https://omnetpp.org>.

Passito, A., Mota, E., Aguiar, R., Carvalho, L.S.G., Moura, E., Nascimento, E.,
Biris, I., Mota, E. 2005, ‘Evaluating voice speech quality in 802.11b networks
with VPN / IPsec’. In Proceedings of the 13th IEEE International Conference on
Networks, Kuala Lumpur: IEEE

Physiome project 2015, JSim, Physiome. Viewed on 7 September, 2015

<http://www.physiome.org/jsim>

Powell, J.M. 2010, ‘The impact of virtual private network (VPN) on a company’s
network’, Research report for Honour’s degree, Utah State University

Pruthi, S. & Pruthi, G. 2009, ‘Performance evaluation of wireless IPsec VPN’.

International Journal of Information Technology and Knowledge Management,
vol. 2, no. 1, pp. 41 – 44

Rahimi, M.R., Hashim, H. & Rahman, R. 2009, ‘implementation of quality of

service (QoS) in multi-protocol label switching (MPLS) networks’. In the 5th

57
International Colloquium on Signal Processing and its Applications, CSPA.
Kuala Lumpur: IEEE

Riverbed, 2012, Opnet, Riverbed. Viewed on 9 September, 2015

<http://za.riverbed.com/products/performance-
managementcontrol/opnet.html?redirect=opnet>

Ross, SM. 2010, Introductory Statistics, 3rd Edition. British Library, Canada

Tan, NK. 2003, Building VPNs with IPsec and MPLS, New York, McGraw-Hill
networking, USA

58
Appendices

Appendix A: Tcl script for scenario 1

##########TCL SCRIPT#####################################
set ns [new Simulator]
set mytracefilescen1 [open out_tracescen1.tr w]
set mynamtracefilescen1 [open out_namtracescen1.nam w]
set frputfile_udp [open out_frputscen1_udp.tr w]
set frputfile_tcp [open out_frputscen1_tcp.tr w]
set frputpktlost_udp [open out_lostpktscen1_udp.tr w]
set frputpktlost_tcp [open out_lostpktscen1_tcp.tr w]
$ns trace-all $mytracefilescen1
$ns namtrace-all $mynamtracefilescen1
#Closing the trace files
proc finish {} {
global ns frputpktlost_udp frputpktlost_tcp frputfile_udp frputfile_tcp
mytracefilescen1 mynamtracefilescen1
$ns flush-trace
close $frputpktlost_udp
close $frputpktlost_tcp
close $frputfile_udp
close $frputfile_tcp
close $mytracefilescen1
close $mynamtracefilescen1
#Call xgraph to display the results
#exec nam out_namtracescen1.nam &

#exec xgraph out_frputscen1_udp.tr -t "Udp Throughput" &

#exec xgraph out_frputscen1_tcp.tr -t "Tcp Throughput" &
#exec xgraph out_lostpktscen1_udp.tr -t "Udp Lost" &

exec awk -f delay.awk out_tracescen1.tr > delayfilescen1.dat

59
 #exec xgraph out_lostpktscen1_tcp.tr -t "Tcp Lost" &
exit 0
}

#
# make nodes
set node1 [$ns node]
set node4 [$ns node]

# make links
$ns duplex-link $node1 $node4 100Mb 10ms DropTail

# Creation of sending agent

set udp [new Agent/UDP]
set tcp [new Agent/TCP]

#Creation of receiving agent / Sink

set nullsink [new Agent/Null]
set tcpsink [new Agent/TCPSink]
set lostpktsink_udp [new Agent/LossMonitor]
set lostpktsink_tcp [new Agent/LossMonitor]

# Creation of application (This is a modelling of users demand):

set cbr [new Application/Traffic/CBR]
$cbr set rate_ 10mb
#$cbr set rate_ 100mb
#$cbr set rate_ 1000mb

$cbr set packetSize_ 512

#$cbr set packetSize_ 1024
#$cbr set packetSize_ 2048
$cbr set interval_ 0.5
#$cbr set random_ false

60
set ftp [new Application/FTP]
$tcp set window_ 10
#$tcp set window_ 50
#$tcp set window_ 100

$tcp set packetSize_ 512

#$tcp set packetSize_ 1024
#$tcp set packetSize_ 2048
#Attaching the agent to an application(This is to create a connection between
an#application and a sending agent):
$cbr attach-agent $udp
$ftp attach-agent $tcp
#Attaching the agent to a low-level network(This is to create a connection betwe#en each
agent and a node):
$ns attach-agent $node1 $udp
$ns attach-agent $node1 $tcp
#$ns attach-agent $node4 $nullsink
$ns attach-agent $node4 $lostpktsink_udp
$ns attach-agent $node4 $tcpsink
#$ns attach-agent $node4 $lostpktsink_tcp

# Associating a sending agent with a receiving agent.

#$ns connect $udp $nullsink
$ns connect $udp $lostpktsink_udp
$ns connect $tcp $tcpsink
#$ns connect $tcp $lostpktsink_tcp

#traffic sinks nullsink/tcpsink/lostpktsink_udp and writes it to the files

#frputfile_udp/frputfile_tcp/ frputpktlost_udp/ frputpktlost_tcp

proc record {} {
global ns nullsink tcpsink lostpktsink_udp lostpktsink_tcp frputfile_udp
frputfile_tcp frputpktlost_udp frputpktlost_tcp defaultRNG

#Get an instance of the simulator

61
 set ns [Simulator instance]
#Set the time after which the procedure should be called again
set time 0.5
#How many bytes have been received by the traffic sinks?
#set bwidth_udp [$nullsink set bytes_]
set bwidth_udp [$lostpktsink_udp set bytes_]
set bwidth_tcp [$tcpsink set bytes_]
#set bwidth_tcp [$lostpktsink_tcp set bytes_]

#How many bytes have been lost?

set lostpkt_udp [$lostpktsink_udp set nlost_]
set lostpkt_tcp [$lostpktsink_tcp set nlost_]

#Get the current time

set now [$ns now]

#Calculate the throughput (in MBit/s): and write it to the files

puts $frputfile_udp "$now [expr $bwidth_udp/$time*8/1000000]"
puts $frputfile_tcp "$now [expr $bwidth_tcp/$time*8/1000000]"
puts $frputpktlost_udp "$now [expr $lostpkt_udp/1]"
#puts $frputpktlost_tcp "$now [expr $lostpkt_tcp/1]"

#Reset the bytes_ values on the traffic sinks

#$nullsink set bytes_ 0
$tcpsink set bytes_ 0
$lostpktsink_udp set bytes_ 0
#$lostpktsink_tcp set bytes_ 0
$lostpktsink_udp set nlost_ 0
#$lostpktsink_tcp set nlost_ 0

#Re-schedule the procedure

$ns at [expr $now+$time] "record"
}

62
# http-packmime.tcl
# useful constants
set CLIENT 0
set SERVER 1

# removes all packet headers

remove-all-packet-headers
# adds TCP/IP headers
add-packet-header IP TCP
# use the Heap scheduler
$ns use-scheduler Heap

# SETUP PACKMIME
set rate 15
set pm [new PackMimeHTTP]
$pm set-client $node1
$pm set-server $node4
# new connections per second
$pm set-rate $rate
# use HTTP/1.1
$pm set-http-1.1

# SETUP PACKMIME RANDOM VARIABLES

#global defaultRNG
# create RNGs (appropriate RNG seeds are assigned automatically)
set flowRNG [new RNG]
set reqsizeRNG [new RNG]
set rspsizeRNG [new RNG]

# create RandomVariables
set flow_arrive [new RandomVariable/PackMimeHTTPFlowArrive $rate]
set req_size [new RandomVariable/PackMimeHTTPFileSize $rate $CLIENT]
set rsp_size [new RandomVariable/PackMimeHTTPFileSize $rate $SERVER]

63
# assign RNGs to RandomVariables
$flow_arrive use-rng $flowRNG
$req_size use-rng $reqsizeRNG
$rsp_size use-rng $rspsizeRNG

# set PackMime variables

$pm set-flow_arrive $flow_arrive
$pm set-req_size $req_size
$pm set-rsp_size $rsp_size
# record HTTP statistics
$pm set-outfile "http_1.dat"

# Scheduling events
$ns at 0.0 "record"
$ns at 2 "$cbr start"
$ns at 7 "$cbr stop"
$ns at 8 "$ftp start"
$ns at 13 "$ftp stop"
$ns at 14 "$pm start"
$ns at 19 "$pm stop"
$ns at 20 "finish"
$ns run
########################################################

64
Appendix B: Tcl script for scenario 2

##########TCL SCRIPT###################################
set ns [new Simulator]
set mytracefilescen2 [open out_tracescen2.tr w]
set mynamtracefilescen2 [open out_namtracescen2.nam w]
set frputfile_udp [open out_frputscen2_udp.tr w]
set frputfile_tcp [open out_frputscen2_tcp.tr w]
set frputpktlost_udp [open out_lostpktscen2_udp.tr w]
set frputpktlost_tcp [open out_lostpktscen2_tcp.tr w]
$ns trace-all $mytracefilescen2
$ns namtrace-all $mynamtracefilescen2
#Closing the trace files
proc finish {} {
global ns frputpktlost_udp frputpktlost_tcp frputfile_udp frputfile_tcp
mytracefilescen2 mynamtracefilescen2
$ns flush-trace
close $frputpktlost_udp
close $frputpktlost_tcp
close $frputfile_udp
close $frputfile_tcp
close $mytracefilescen2
close $mynamtracefilescen2
#Call xgraph to display the results
#exec nam out_namtracescen2.nam &

#exec xgraph out_frputscen2_udp.tr -t "Udp Throughput" &

#exec xgraph out_frputscen2_tcp.tr -t "Tcp Throughput" &
#exec xgraph out_lostpktscen2_udp.tr -t "Udp Lost" &

exec awk -f delay.awk out_tracescen2.tr > delayfilescen2.dat

#exec xgraph out_lostpktscen2_tcp.tr -t "Tcp Lost" &

exit 0
}

65
# routing protocol Distance Vector
$ns rtproto DV
#
# make nodes
set node1 [$ns node]
set node2 [$ns node]
set node3 [$ns node]
set node4 [$ns node]

# make links
$ns duplex-link $node1 $node2 100Mb 10ms DropTail
$ns duplex-link $node2 $node3 100Mb 10ms DropTail
$ns duplex-link $node3 $node4 100Mb 10ms DropTail

# Creation of sending agent

set udp [new Agent/UDP]
set tcp [new Agent/TCP]

#Creation of receiving agent / Sink

set nullsink [new Agent/Null]
set tcpsink [new Agent/TCPSink]
set lostpktsink_udp [new Agent/LossMonitor]
set lostpktsink_tcp [new Agent/LossMonitor]

# Creation of application (This is a modelling of users demand):

set cbr [new Application/Traffic/CBR]
$cbr set rate_ 10mb
#$cbr set rate_ 100mb
#$cbr set rate_ 1000mb

$cbr set packetSize_ 512

#$cbr set packetSize_ 1024
#$cbr set packetSize_ 2048
$cbr set interval_ 0.5

66
#$cbr set random_ false

set ftp [new Application/FTP]

$tcp set window_ 10
#$tcp set window_ 50
#$tcp set window_ 100

$tcp set packetSize_ 512

#$tcp set packetSize_ 1024
#$tcp set packetSize_ 2048
#Attaching the agent to an application(This is to create a connection between
an#application and a sending agent):
$cbr attach-agent $udp
$ftp attach-agent $tcp
#Attaching the agent to a low-level network(This is to create a connection betwe#en each
agent and a node):
$ns attach-agent $node1 $udp
$ns attach-agent $node1 $tcp
#$ns attach-agent $node4 $nullsink
$ns attach-agent $node4 $lostpktsink_udp
$ns attach-agent $node4 $tcpsink
#$ns attach-agent $node4 $lostpktsink_tcp

# Associating a sending agent with a receiving agent.

#$ns connect $udp $nullsink
$ns connect $udp $lostpktsink_udp
$ns connect $tcp $tcpsink
#$ns connect $tcp $lostpktsink_tcp

#traffic sinks nullsink/tcpsink/lostpktsink_udp and writes it to the files

#frputfile_udp/frputfile_tcp/ frputpktlost_udp/ frputpktlost_tcp

proc record {} {
global ns nullsink tcpsink lostpktsink_udp lostpktsink_tcp frputfile_udp
frputfile_tcp frputpktlost_udp frputpktlost_tcp defaultRNG

67
 #Get an instance of the simulator
set ns [Simulator instance]
#Set the time after which the procedure should be called again
set time 0.5
#How many bytes have been received by the traffic sinks?
#set bwidth_udp [$nullsink set bytes_]
set bwidth_udp [$lostpktsink_udp set bytes_]
set bwidth_tcp [$tcpsink set bytes_]
#set bwidth_tcp [$lostpktsink_tcp set bytes_]

#How many bytes have been lost?

set lostpkt_udp [$lostpktsink_udp set nlost_]
set lostpkt_tcp [$lostpktsink_tcp set nlost_]

#Get the current time

set now [$ns now]
#Calculate the throughput (in MBit/s): and write it to the files
puts $frputfile_udp "$now [expr $bwidth_udp/$time*8/1000000]"
puts $frputfile_tcp "$now [expr $bwidth_tcp/$time*8/1000000]"
puts $frputpktlost_udp "$now [expr $lostpkt_udp/1]"
#puts $frputpktlost_tcp "$now [expr $lostpkt_tcp/1]"

#Reset the bytes_ values on the traffic sinks

#$nullsink set bytes_ 0
$tcpsink set bytes_ 0
$lostpktsink_udp set bytes_ 0
#$lostpktsink_tcp set bytes_ 0
$lostpktsink_udp set nlost_ 0
#$lostpktsink_tcp set nlost_ 0

#Re-schedule the procedure

$ns at [expr $now+$time] "record"
}
# http-packmime.tcl

68
# useful constants
set CLIENT 0
set SERVER 1

# removes all packet headers

remove-all-packet-headers
# adds TCP/IP headers
add-packet-header IP TCP
# use the Heap scheduler
$ns use-scheduler Heap
# SETUP PACKMIME
set rate 15
set pm [new PackMimeHTTP]
$pm set-client $node1
$pm set-server $node4
# new connections per second
$pm set-rate $rate
# use HTTP/1.1
$pm set-http-1.1

# SETUP PACKMIME RANDOM VARIABLES

#global defaultRNG

# create RNGs (appropriate RNG seeds are assigned automatically)

set flowRNG [new RNG]
set reqsizeRNG [new RNG]
set rspsizeRNG [new RNG]

# create RandomVariables
set flow_arrive [new RandomVariable/PackMimeHTTPFlowArrive $rate]
set req_size [new RandomVariable/PackMimeHTTPFileSize $rate $CLIENT]
set rsp_size [new RandomVariable/PackMimeHTTPFileSize $rate $SERVER]
# assign RNGs to RandomVariables
$flow_arrive use-rng $flowRNG
$req_size use-rng $reqsizeRNG

69
$rsp_size use-rng $rspsizeRNG

# set PackMime variables

$pm set-flow_arrive $flow_arrive
$pm set-req_size $req_size
$pm set-rsp_size $rsp_size

# record HTTP statistics

$pm set-outfile "http_2.dat"
# Scheduling events
$ns at 0.0 "record"
$ns at 2 "$cbr start"
$ns at 7 "$cbr stop"
$ns at 8 "$ftp start"
$ns at 13 "$ftp stop"
$ns at 14 "$pm start"
$ns at 19 "$pm stop"
$ns at 20 "finish"
$ns run
#######################################################

70
Appendix C: Awk script for delay

##########AWK SCRIPT#################################
BEGIN {
# simple awk script to generate end-to-end packet lifetime statistics
# in a form suitable for plotting with xgraph.
# Lloyd Wood, July 1999.
# http://www.ee.surrey.ac.uk/Personal/L.Wood/ns/
highest_packet_id = 0;
}
{
action = $1;
time = $2;
node_1 = $3;
node_2 = $4;
src = $5;
flow_id = $8;
node_1_address = $9;
node_2_address = $10;
seq_no = $11;
packet_id = $12;
if ( packet_id > highest_packet_id ) highest_packet_id = packet_id;

# getting start time is not a problem, provided you're not starting

# traffic at 0.0.
# could test for sending node_1_address or flow_id here.
if ( start_time[packet_id] == 0 ) start_time[packet_id] = time;

# only useful for small unicast where packet_id doesn't wrap.

# checking receive means avoiding recording drops
if ( action != "d" ) {
if ( action == "r" ) {
# could test for receiving node_2_address or flow_id here.
end_time[packet_id] = time;

71
}
} else {
end_time[packet_id] = -1;
}
}
END {
for ( packet_id = 0; packet_id <= highest_packet_id; packet_id++ ) {
start = start_time[packet_id];
end = end_time[packet_id];
packet_duration = end - start;
if ( start < end ) printf("%d %f\n", start, packet_duration);
}
}
#############################################

72
Appendix D: HTTP throughput

Connection rate 5
With
No VPN VPN
427 10410 220.45 49.15854 427 10410 460.974 23.50892
479 182 57.414 11.51287 479 182 137.545 4.8057
571 1167 82.703 21.01496 510 134 195.826 3.288634
510 134 115.697 5.566264 571 1167 242.946 7.153853
874 1375 180.212 12.47975 874 1375 340.537 6.604275
479 182 43.033 15.3603 479 182 83.151 7.949393
3804 14653 186.299 99.07192 470 1202 182.587 9.157278
479 182 22.267 29.68518 479 182 160.178 4.126659
470 1202 160.159 10.43963 3804 14653 547.052 33.73902
733 6066 122.694 55.41428 441 419 220.248 3.904689
441 419 140.085 6.13913 733 6066 363.173 18.72111
733 2895 74.925 48.42176 317 630 187.309 5.055817
733 2940 63.782 57.58678 386 1028 199.524 7.086867
431 7376 119.631 65.259 733 2895 244.11 14.86215
317 630 65.683 14.41773 431 7376 315.566 24.73967
386 1028 67.156 21.05545 733 2940 246.25 14.91574
474 7764 101.483 81.17616 474 7764 301.922 27.28519
869 175 61.439 16.99246 869 175 181.645 5.747474
445 248 140.071 4.947491 445 248 220.208 3.147025
417 17562 141.258 127.2777 343 493 174.615 4.787676
1013 6731 123.979 62.46219 417 17562 421.908 42.61356
343 493 60.427 13.83488 1013 6731 364.523 21.2442
417 701 94.475 11.83382 417 701 140.651 7.948753
417 572 47.483 20.82851 441 609 187.704 5.593914
441 609 62.722 16.74054 417 572 122.892 8.047717
278 938 160.12 7.594304 278 938 280.353 4.337389
417 775 43.328 27.51108 491 594 189.24 5.73346
491 594 63.204 17.16664 396 2647 243.819 12.48057
396 2647 89.152 34.13272 452 1944 243.621 9.834949
452 1944 83.112 28.82857 447 3460 253.813 15.39322
447 3460 82.579 47.31227 488 5168 303.711 18.62297
488 5168 113.494 49.83523 428 7503 303.264 26.15213
428 7503 103.172 76.87163 447 4898 123.315 43.34428
447 4898 46.955 113.8324 417 775 126.813 9.399667
447 11330 87.254 134.9738 488 125 128.237 4.780212
447 3576 48.399 83.12155 467 202 120.422 5.555463
488 125 42.091 14.56368 447 11330 207.547 56.74377
447 9922 41.117 252.1828 418 133 122.08 4.513434
467 202 41.975 15.93806 443 329 121.185 6.370425

73
 860 482 61.511 21.81724 409 185 127.623 4.654333
418 133 96.299 5.721762 860 482 280.383 4.78631
443 329 47.517 16.24682 447 3576 121.973 32.98271
831 2665 200.275 17.456 831 2665 356.966 9.79365
448 123269 322.571 383.5342 447 9922 221.123 46.89245
409 185 140.064 4.240918 448 123269 965.201 128.1774
831 9174 217.546 45.99027 831 9174 458.108 21.83983
831 3172 82.283 48.64917

Connection rate 10
With
No VPN VPN
427 10410 220.45 49.15854 494 134 122.724 5.117173
494 134 42.598 14.74248 478 182 137.545 4.79843
478 182 57.414 11.49545 427 10410 460.974 23.50892
534 1167 135.792 12.52651 534 1167 256.016 6.644116
860 1375 180.211 12.40213 860 1375 340.534 6.563221
3492 14653 184.013 98.60716 478 182 62.385 10.57947
478 182 22.267 29.64027 461 1202 185.744 8.953183
461 1202 65.525 25.37963 3492 14653 504.703 35.95184
478 182 22.327 29.56062 438 419 220.247 3.891086
438 419 140.085 6.117714 478 182 160.178 4.120416
743 6066 220.428 30.88991 743 6066 363.175 18.74854
377 1028 63.591 22.09432 377 1028 183.854 7.641933
312 630 79.371 11.86831 312 630 199.41 4.723936
427 7376 115.156 67.76026 427 7376 315.565 24.72708
743 2895 87.301 41.67192 473 116 125.768 4.683226
743 2940 65.912 55.87753 743 2895 247.677 14.68849
473 116 41.001 14.3655 743 2940 181.549 20.28653
823 20152 161.97 129.4993 416 168 121.065 4.823855
416 168 40.946 14.26269 432 242 220.203 3.060812
416 168 23.527 24.82254 416 168 63.633 9.177628
432 242 140.07 4.81188 823 20152 482.684 43.45493
416 168 27.444 21.2797 1167 105 200.628 6.340092
1167 105 114.5 11.10917 416 168 67.55 8.645448
337 938 80.469 15.84461 337 938 234.739 5.431564
448 2647 82.853 37.35532 448 2647 243.227 12.72474
283 594 160.093 5.478066 283 594 280.271 3.129114
405 1944 83.394 28.16749 493 112 123.121 4.913865
493 112 42.998 14.07042 405 1944 243.735 9.637516
448 3460 83.403 46.85683 448 3460 243.736 16.03374
492 5168 109.242 51.81157 492 5168 309.697 18.27593
452 9474 133.542 74.32868 452 9474 374.092 26.53358
428 13717 123.005 114.9953 416 168 62.814 9.29729

74
 416 168 22.708 25.71781 428 13717 363.544 38.90863
452 701 46.594 24.74568 410 482 120.476 7.403964
452 572 48.143 21.26997 445 132 128.228 4.499797
445 132 42.088 13.70937 768 329 182.23 6.019865
410 482 41.993 21.24164 433 554 182.324 5.41344
768 329 61.068 17.96358 452 701 121.273 9.507475
454 4885 100.517 53.11539 829 185 236.543 4.286747
387 119 41.438 12.21101 454 4885 327.762 16.28926
433 554 116.34 8.483755 387 119 220.163 2.298297
452 775 47.55 25.80442 452 572 121.666 8.416484
829 185 160.104 6.333383 1079 256 181.216 7.366899
474 123269 348.677 354.8929 429 2691 247.974 12.58196
452 1386 140.163 13.1133 403 1100 182.445 8.238099
429 2691 82.216 37.94882 433 1671 340.456 6.179947
433 1671 81.072 25.95224 768 10856 363.051 32.01754
1079 256 99.74 13.3848 676 3011 310.489 11.87482
768 10856 217.685 53.39826 452 775 220.333 5.568844
450 22862 190.035 122.6721 285 4555 340.272 14.22391
403 1100 61.017 24.63248 450 22862 518.652 44.94729
676 3011 110.092 33.49017 449 27103 471.039 58.49197
449 27103 143.305 192.2613 452 1386 121.268 15.15651
285 4555 200.254 24.1693 474 123269 969.679 127.6123
768 3861 101.68 45.52518 428 4196 302.144 15.30396
11838 1665 260.642 51.80669 768 3861 302.643 15.29525
428 4196 102.184 45.2517 11838 1665 581.276 23.22993
321 721 62.716 16.61458 321 721 182.921 5.696448
563 2889 100.668 34.29094 276 309 121.828 4.801852
276 309 41.709 14.02575 563 2889 301.027 11.46741
2125 174 82.494 27.86869 2125 174 244.064 9.41966
904 720 83.719 19.39823 904 720 242.794 6.688798
439 142 46.522 12.48872 439 142 126.64 4.587808
352 10524 142.498 76.32388 277 576 183.663 4.644376
1649 463 83.552 25.27767 448 67 220.165 2.339155
448 67 140.057 3.677074 352 10524 383.028 28.39479
277 576 160.091 5.32822 1649 463 340.458 6.203408
1649 3138 102.366 46.76357 277 3129 182.52 18.66097
277 3129 61.362 55.50667 364 1455 181.614 10.01575
364 1455 61.76 29.45272 552 1013 242.044 6.465767
552 1013 126.602 12.36157 1649 3138 287.215 16.66696
1649 1383 81.27 37.30774 893 109 182.302 5.496374
277 6883 102.481 69.86661 275 1326 181.404 8.825605
275 1326 62.154 25.7586 277 6883 262.906 27.23407
893 109 93.084 10.76447 372 5303 333.813 17.00054
655 6262 135.314 51.11814 655 6262 375.812 18.40548

75
 351 7117 107.405 69.53121 1649 1383 242.891 12.48297
372 5303 200.324 28.32911 351 7117 400.882 18.62892
826 19455 162.876 124.518 826 19455 488.191 41.54317
373 578 41.255 23.05175 298 990 190.185 6.772353
431 773 60.639 19.85521 948 23801 489.438 50.56616
407 892 69.941 18.5728 373 578 126.553 7.514638
298 990 66.412 19.39409 398 245231 1784.324 137.6594

Connection rate 15 With

No VPN VPN
487 134 42.597 14.57849 487 134 122.722 5.060217
476 182 57.414 11.46062 476 182 137.545 4.783889
427 10410 220.45 49.15854 517 1167 256.012 6.577817
517 1167 135.79 12.4015 427 10410 460.974 23.50892
851 1375 180.21 12.35226 851 1375 340.532 6.536831
456 1202 65.525 25.30332 456 1202 203.35 8.15343
3301 14653 183.998 97.57715 436 419 122.455 6.982157
476 182 22.267 29.55046 476 182 160.177 4.107956
436 419 42.352 20.18795 3301 14653 487.051 36.86267
476 182 39.183 16.793 310 630 183.743 5.115841
310 630 63.554 14.79057 373 1028 195 7.184615
746 6066 220.428 30.90351 746 6066 363.176 18.75675
373 1028 74.737 18.74574 476 182 79.301 8.297499
424 7376 200.507 38.90138 472 116 127.203 4.622532
746 2895 87.301 41.70628 424 7376 400.961 19.45326
472 116 45.648 12.88118 746 2895 246.242 14.78627
746 2940 61.265 60.16486 427 242 121.536 5.504542
789 20152 161.967 129.2918 746 2940 181.495 20.3091
416 168 40.946 14.26269 416 168 220.182 2.652351
416 168 23.527 24.82254 334 938 183.84 6.91906
427 242 140.07 4.776183 1278 105 234.683 5.893056
334 938 80.469 15.80733 416 168 67.55 8.645448
1278 105 114.509 12.07765 789 20152 482.281 43.42074
416 168 27.444 21.2797 287 594 182.888 4.817156
452 2647 82.853 37.40359 452 2647 260.957 11.87552
287 594 160.093 5.503051 416 168 160.159 3.646376
495 112 42.998 14.11694 495 112 123.121 4.930109
410 1944 83.394 28.22745 410 1944 243.736 9.657991
446 3460 83.403 46.83285 446 3460 243.736 16.02554
492 5168 109.242 51.81157 492 5168 309.697 18.27593
455 9474 133.542 74.35114 406 482 128.556 6.907496
428 13717 123.005 114.9953 443 132 147.241 3.905162
455 701 42.76 27.03461 455 9474 374.092 26.5416

76
 406 482 40.307 22.03091 428 13717 363.544 38.90863
443 132 48.11 11.95178 727 329 182.658 5.781296
727 329 62.133 16.9958 808 185 181.257 5.478409
808 185 61.06 16.26269 377 119 122.066 4.063376
429 554 62.007 15.85305 429 554 180.517 5.445471
449 4885 127.358 41.88194 449 4885 307.188 17.36396
416 168 21.435 27.24516 455 701 176.554 6.547572
455 572 96.333 10.66094 431 1671 241.837 8.691805
377 119 140.056 3.541441 403 1100 181.257 8.292094
425 2691 87.681 35.53792 1043 256 280.373 4.633114
1043 256 61.044 21.27973 425 2691 340.534 9.150334
467 123269 328.117 377.1094 455 572 122.494 8.384084
727 10856 122.483 94.56823 283 4555 302.637 15.98615
455 775 89.313 13.77179 678 3011 340.299 10.84047
403 1100 69.944 21.48862 449 22862 428.919 54.34826
431 1671 180.154 11.6678 416 168 160.159 3.646376
283 4555 102.36 47.26456 727 10856 410.269 28.2327
449 22862 238.005 97.94332 455 775 121.204 10.14818
455 1386 41.034 44.86523 442 27103 519.007 53.0725
442 27103 180.806 152.3456 424 4196 310.492 14.87961
678 3011 200.294 18.41793 12280 1665 482.832 28.88168
727 3861 101.677 45.12328 273 309 120.575 4.826871
424 4196 102.184 45.21256 467 123269 965.88 128.107
12280 1665 260.678 53.49512 455 1386 122.348 15.04724
318 721 62.715 16.56701 318 721 280.31 3.706611
553 2889 100.668 34.1916 553 2889 303.241 11.35071
273 309 41.709 13.95382 2088 174 244.055 9.268403
2088 174 83.727 27.01637 887 720 242.112 6.637424
887 720 82.481 19.48327 434 142 126.639 4.548362
434 142 46.521 12.38151 727 3861 303.006 15.14161
344 10524 142.497 76.26827 265 576 183.66 4.579114
439 67 43.454 11.6445 439 67 220.163 2.298297
1515 463 160.184 12.3483 1515 463 280.453 7.052875
265 576 160.09 5.253295 344 10524 383.026 28.37405
1515 3138 82.395 56.47187 356 1455 181.612 9.971808
265 3129 61.361 55.31201 265 3129 182.517 18.59553
540 1013 81.742 18.9988 540 1013 286.901 5.413017
356 1455 106.619 16.98572 1515 3138 242.36 19.19871
262 1326 61.194 25.95026 262 1326 181.401 8.754086
876 109 82.139 11.99187 876 109 202.335 4.868164
1515 1383 62.258 46.54823 371 5303 302.782 18.73956
346 7117 115.358 64.69426 334 805 187.259 6.082485
371 5303 133.305 42.56404 346 7117 333.861 22.35361
265 6883 87.366 81.81673 670 1124 244.802 7.328372

77
651 6262 220.436 31.36058 265 6883 243.163 29.39592
334 805 81.281 14.01311 651 6262 460.933 14.99784
828 19455 162.876 124.5303 369 2378 261.695 10.49695
670 1124 84.55 21.21821 1515 1383 235.924 12.28362
369 2378 135.545 20.26633 828 19455 496.145 40.88119
427 688 60.494 18.43158 427 688 280.328 3.977484
346 7160 107.148 70.05264 346 7160 301.273 24.91428
427 559 48.115 20.49257 1436 4995 367.564 17.49627
1436 4995 220.394 29.17956 427 559 121.177 8.136858
427 682 41.003 27.0468 407 2150 248.465 10.29119
407 2150 80.519 31.75648 278 150 121.349 3.527017
886 4182 138.827 36.50587 443 996 198.957 7.232719
2871 2154 124.873 40.24088 377 3652 245.177 16.43303
377 3652 81.337 49.53465 886 4182 361.146 14.03311
278 150 41.255 10.3745 2871 2154 361.839 13.88739
443 996 160.138 8.986 427 682 220.305 5.03393
877 3865 220.249 21.53018 877 3865 460.746 10.292
425 21202 141.216 153.1484 319 370 122.202 5.638206
390 18715 145.743 131.0869 425 21202 421.85 51.26704
2871 12265 152.817 99.04657 390 18715 426.403 44.80503
425 2440 83.914 34.1421 347 9829 372.945 27.28552
347 9829 220.394 46.17186 425 2440 340.556 8.412713
319 370 41.903 16.44274 498 138 122.857 5.17675
443 3453 82.302 47.33785 443 3453 244.334 15.94539
469 1311 61.213 29.07879 469 1311 187.007 9.51836
498 138 42.729 14.8845 466 142 121.63 4.998767
428 11631 127.155 94.83701 434 707 182.524 6.251233
1041 7664 120.87 72.01953 943 142 185.733 5.841719
434 707 62.303 18.31372 1799 104 241.421 7.882496
943 142 65.055 16.6782 428 11631 363.102 33.21105
1799 104 85.55 22.2443 2871 12265 422.525 35.82273
466 142 41.49 14.65413 434 2018 245.486 9.98835
426 16577 214.507 79.26548 1041 7664 361.396 24.08715
434 2018 81.708 30.0093 247 207 122.845 3.695714
331 11370 124.462 94.01263 390 2493 241.986 11.91391
390 2493 81.235 35.48963 426 16577 455.075 37.36307
196 4348 101.419 44.80423 331 11370 364.953 32.06166
684 268 62.793 15.16093 378 463 129.554 6.491502
398 245231 661.628 371.2494 684 268 205.689 4.628347
378 463 41.002 20.51119 418 459 121.171 7.237705
247 207 65.452 6.936381 429 600 180.619 5.697075
714 37844 186.192 207.0873 196 4348 301.782 15.05723
429 600 69.415 14.82389 1411 466 183.814 10.21141
418 459 47.846 18.32964 447 1544 185.694 10.72194

78
1411 466 66.096 28.39809 528 639 186.261 6.265402
447 1544 65.423 30.43272 373 7261 305.637 24.97734
948 23801 161.272 153.4612 431 773 181.627 6.62897
528 639 63.501 18.37766 407 892 180.893 7.181041
373 7261 105.209 72.56033 409 10496 362.977 30.04323
409 10496 122.442 89.06258 714 37844 546.988 70.49149
373 578 41.255 23.05175 298 990 190.185 6.772353
431 773 60.639 19.85521 948 23801 489.438 50.56616
407 892 69.941 18.5728 373 578 126.553 7.514638
298 990 66.412 19.39409 398 245231 1784.324 137.6594

79
Appendix E: Instant CBR throughput
CBR transfer rate: 10 Mb
Packet size: 512 Bytes Packet size: 1024 Bytes Packet size: 2048 Bytes
Time WithoutVpn WithVpn Time WithoutVpn WithVpn Time WithoutVpn WithVpn
2 0 0 2 0 0 2 0 0
2.5 0.008 0.008 2.5 0.016 0.016 2.5 0.033 0.033
3 0.008 0.008 3 0.016 0.016 3 0.033 0.033
3.5 0.008 0.008 3.5 0.016 0.016 3.5 0.033 0.033
4 0.008 0.008 4 0.016 0.016 4 0.033 0.033
4.5 0.008 0.008 4.5 0.016 0.016 4.5 0.033 0.033
5 0.008 0.008 5 0.016 0.016 5 0.033 0.033
5.5 0.008 0.008 5.5 0.016 0.016 5.5 0.033 0.033
6 0.008 0.008 6 0.016 0.016 6 0.033 0.033
6.5 0.008 0.008 6.5 0.016 0.016 6.5 0.033 0.033
7 0.008 0.008 7 0.016 0.016 7 0.033 0.033
CBR transfer rate: 100 Mb
Packet size: 512 Bytes Packet size: 1024 Bytes Packet size: 2048 Bytes
Time WithoutVpn WithVpn Time WithoutVpn WithVpn Time WithoutVpn WithVpn
2 0 0 2 0 0 2 0 0
2.5 0.008 0.008 2.5 0.016 0.016 2.5 0.033 0.033
3 0.008 0.008 3 0.016 0.016 3 0.033 0.033
3.5 0.008 0.008 3.5 0.016 0.016 3.5 0.033 0.033
4 0.008 0.008 4 0.016 0.016 4 0.033 0.033
4.5 0.008 0.008 4.5 0.016 0.016 4.5 0.033 0.033
5 0.008 0.008 5 0.016 0.016 5 0.033 0.033
5.5 0.008 0.008 5.5 0.016 0.016 5.5 0.033 0.033
6 0.008 0.008 6 0.016 0.016 6 0.033 0.033
6.5 0.008 0.008 6.5 0.016 0.016 6.5 0.033 0.033
7 0.008 0.008 7 0.016 0.016 7 0.033 0.033
CBR transfer rate: 1000 Mb
Packet size: 512 Bytes Packet size: 1024 Bytes Packet size: 2048 Bytes
Time WithoutVpn WithVpn Time WithoutVpn WithVpn Time WithoutVpn WithVpn
2 0 0 2 0 0 2 0 0
2.5 0.008 0.008 2.5 0.016 0.016 2.5 0.033 0.033
3 0.008 0.008 3 0.016 0.016 3 0.033 0.033
3.5 0.008 0.008 3.5 0.016 0.016 3.5 0.033 0.033
4 0.008 0.008 4 0.016 0.016 4 0.033 0.033
4.5 0.008 0.008 4.5 0.016 0.016 4.5 0.033 0.033
5 0.008 0.008 5 0.016 0.016 5 0.033 0.033
5.5 0.008 0.008 5.5 0.016 0.016 5.5 0.033 0.033
6 0.008 0.008 6 0.016 0.016 6 0.033 0.033
6.5 0.008 0.008 6.5 0.016 0.016 6.5 0.033 0.033
7 0.008 0.008 7 0.016 0.016 7 0.033 0.033

80
Appendix F: Instant FTP throughput

APPENDIX F: Instant FTP FTP window size: 10 Kb

throughput
Packet size: 512 Packet size: 1024 Packet size: 2048
Time WithoutVpn WithVpn Time WithoutVpn WithVpn Time WithoutVpn WithVpn
8 0 0 8 0 0 8 0 0
8.5 1.979 0.48 8.5 3.81 0.92 8.5 7.48 1.8
9 2.208 0.79 9 4.26 1.53 9 8.35 3.01
9.5 2.208 0.71 9.5 4.26 1.36 9.5 8.02 2.67
10 2.208 0.71 10 4.26 1.36 10 8.35 2.67
10.5 2.208 0.79 10.5 4.09 1.36 10.5 8.35 2.67
11 2.208 0.71 11 4.26 1.53 11 8.35 3.01
11.5 2.208 0.71 11.5 4.26 1.36 11.5 8.15 2.67
12 2.208 0.79 12 4.26 1.36 12 8.22 2.67
12.5 2.12 0.71 12.5 4.26 1.53 12.5 8.35 2.67
13 0.088 0.71 13 4.26 1.36 13 8.35 3.01
FTP window size: 50 Kb
Packet size: 512 Packet size: 1024 Packet size: 2048
Time WithoutVpn WithVpn Time WithoutVpn WithVpn Time WithoutVpn WithVpn
8 0 0 8 0 0 8 0 0
8.5 8.94 1.43 8.5 17.23 2.76 8.5 33.34 5.41
9 11.04 3.97 9 21.28 7.66 9 40.89 13.73
9.5 11.04 3.53 9.5 21.13 6.81 9.5 41.43 14.67
10 11.04 3.53 10 20.84 6.81 10 41.76 13.36
10.5 11.04 3.97 10.5 21.02 6.81 10.5 41.76 13.36
11 11.04 3.53 11 21.28 7.66 11 41.06 14.37
11.5 10.95 3.53 11.5 21.28 6.81 11.5 40.92 14.03
12 10.8 3.65 12 21.28 6.81 12 41.63 13.36
12.5 10.93 3.85 12.5 21.28 7.66 12.5 41.76 13.36
13 10.04 3.53 13 21.28 6.81 13 41.69 15
FTP window size: 100 Kb
Packet size: 512 Packet size: 1024 Packet size: 2048
Time WithoutVpn WithVpn Time WithoutVpn WithVpn Time WithoutVpn WithVpn
8 0 0 8 0 0 8 0 0
8.5 17.01 1.997 8.5 32.67 3.85 8.5 62.21 7.55
9 22.08 7.95 9 42.12 14.78 9 82.65 27.09
9.5 22.08 7.07 9.5 42.12 14.16 9.5 83.19 29.7
10 22.08 7.07 10 42.12 13.62 10 82.85 26.73
10.5 22.03 7.9 10.5 42.3 13.62 10.5 82.68 26.73
11 21.84 7.11 11 42.56 15.32 11 82.65 27.73
11.5 21.84 7.07 11.5 42.56 13.62 11.5 82.68 29.06
12 21.84 7.19 12 42.56 13.62 12 83.35 26.73
12.5 21.97 7.83 12.5 42.56 15.32 12.5 82.68 26.73
13 22.08 7.07 13 42.49 13.62 13 82.65 28.36

81