Comparing the Accuracy of Network Utilization Performance between Real Network and Simulation Model for Local Area Network (LAN)

doi:10.4236/ijcns.2008.14042

I. J. Communications, Network and System Sciences, 2008, 4, 285-385

Published Online November 2008 in SciRes (http://www. SciRP.org/journal/ijcns/)

.

Comparing the Accuracy of Network Utilization

Performance between Real Network and Simulation Model

for Local Area Network (LAN)

Mohd Nazri ISMAIL

1

, Abdullah Mohd ZIN

2

1

Faculty of MIIT, University Kuala Lumpur, Malaysia

2

Faculty of Computer Science, UKM, Malaysia

Email: mnazrii@miit.unikl.edu.my

Received March 6, 2008; revised August 22, 2008; accepted October 8, 2008

Abstract

This article presents a novel approach for the measurement and estimation of network traffic utilization

between network nodes in heterogeneous environment. This research investigates performance evaluation of

network interface on heterogeneous services and technologies environment. This study proposes an enhanced

equation to evaluate the performance of network interface via Little Law and Queuing theories to improve

the evaluation algorithm. To get accuracy results on the performance of simulation model, it measures

(verify and validate) data from Local Area Network (real network environment). This project uses network

management tool to capture those data and Fluke Optiview device to generate traffic. As a result, this

simulation model can provide a good approximation of the real traffic observed in the real network

environment. Through laboratory and field experiments, the result shows that the model via simulation is

capable of approximating the performance of network utilization and traffic over heterogeneous services and

techniques within a minimum error range.

Keywords: Network Utilization, Real Network, LAN

1. Introduction

Considerable research has been conducted to model and

quantify the performance of heterogeneous services and

technologies (e.g., [1

–

3]). Accurate measurements and

analyses of network characteristics are essential for

robust network performance and management. However,

no current research specifically focuses on using queuing

theory to measure heterogeneous services and technologies

performance, which is the object of this research.

Queuing theory [4] has been used as an effective tool to

model performance in several technical and social

contexts. Evaluating the performance of a computer

networking usually involves constructing an appropriate

model to predict the heterogeneous environment

behaviour via simulation model. The heterogeneous

environment model is then analyzed and simulated using

mathematical techniques. For example, several flow-

level network traffic models have been proposed to

describe/stimulate [5

–

7]. These models have been used to

study fairness, response times, queue lengths and loss

probabilities under different assumptions and using a

variety of mathematical techniques. Queuing theory has

been widely used to model and analyze the network

performance of complex systems involving services,

communication systems, computer networks and

vehicular traffic. In contrast to other works in the

literature (e.g., [8

–

10]), developed simulation model to

measure the performance of heterogeneous environment.

Our model can be used to generate representative packet

traffic in a live network environment or in a simulated

environment.

The significant of this study was to develop a

simulation model to measure the performance of network

traffic utilization in heterogeneous network environment

using Queuing theory. This model could assist network

administrators to design and manage heterogeneous

network systems. This simulation model can be used in

various services and technologies to measure heterogeneous

environment. Therefore, this simulation model is designed

340 M. N. ISMAIL ET AL.

to: 1) predict the performance of various services (e.g.

video, audio, voice and message) in order to aid

technology assessment and capacity planning; 2) predict

the expected behavior of new services and designs

through qualitative or quantitative estimates of network

performance; 3) assist network administrator to prepare,

propose, plan and design network topology more

effective and systematic; and 4) conduct “What-If”

analysis for evaluating heterogeneous network

environment performance. Moreover, in the future, the

integration of data and communication services, almost

every “Internet Ready” device will be a communicable

device [11]. With the availability of this infrastructure,

users are now demanding and expecting more services

[12,13]. Convergence is pushing towards an environment

that requires new investment in infrastructure and able to

support the delivery of rich services (various services),

applications and content [5,14]. In addition, more people

are using multimedia services such as MMS, WAP, i-

mode or push-to-talk. GPRS (General Packet Radio

Service) is an overlay on GSM networks that allows this

kind of end-to-end IP-based packet traffic from mobile

devices to the Internet [15]. Network deployment is

growing increasing complex as the industry lashes

together a mix of wired and wireless technologies into

large-scale heterogeneous network architecture and as

user applications and traffic continue to evolve. Faced

with this growing complexity, network designers and

researchers almost universally use simulation in order to

predict the expected performance of complex networks

[16]. The successful evolution of the Internet is tightly

coupled to the ability to design simple and accurate

models [17]. Many factors may contribute to the

congestion of network interface, such as a heavy load in

the network that usually generates higher traffic. Once

the number of requests exceeds the maximum capability

of network, many clients will not able to receive

responses from the network [18]. Thus, this research is

critical to be conducted in order to predict and measure

of network traffic utilization in heterogeneous

environment.

2. Problem Statements

In the 21 century, a network infrastructure is based on

multi-service implementation over convergence of

network medium such as ISP, PSTN and GSM [19,20].

Availability of various services has produced multi-

traffic in network infrastructure. Therefore, multi-traffic

in the network infrastructure has become more complex

to observe and analyze [14,21,22]. Today, retrieving and

sending information can be done using a variety of

technologies such as PC, PDA, fix and mobile phones

via the wireless, high speed network, ISDN and ADSL

lines that are more prone to heterogeneous environment,

but unfortunately the optimal capability of technologies

are not fully realized. The main factors of network

congestion are related to network design and bandwidth

capacity [23]. Nevertheless, few studies have been

conducted to evaluate the application of computer

network technologies and services over heterogeneous

environment especially in Higher Education Institutes.

Algorithms for actively measuring network physical and

available bandwidths have been researched for many

years. Many tools have been developed, and only a few

tools have successfully achieved a close estimation of

network bandwidths [3]. Therefore, retrieving and

sending information over heterogeneous environment

using convergence of technologies in Higher Educational

Institutes should be analyzed and evaluated via

simulation model. This research has setup a pilot test-bed

(real network environment) to analyze and measure of

network traffic utilization at University of Kuala Lumpur

in Malaysia. This study posits several research questions:

1) what is the performance level of the network utili-

zation and traffic; and 2) Is the simulation model for

evaluating and measuring the heterogeneous environment

performance effective?

3. Methodologies

Whatever modeling paradigm or solution techniques in

heterogeneous environment model development are

being used, the performance measures extracted from a

simulation model must be a good representation of the

real network environment. Inevitably, some assumptions

must be made about the real network in order to

construct the heterogeneous environment model. Figure

1 shows the overall framework of the simulation model.

There are four performance techniques to validate the

simulation model: 1) graphical representation; 2) tracing;

3) parameter variability; and 4) predictive validation. In

addition, there are two techniques to judging how good a

model is with respect to the real network: 1) it must

ascertain whether the simulation model implements the

assumptions correctly (model verification); and 2)

assumptions which have been made are reasonable with

respect to the real network (model validation).

Comparison with a real network is the most reliable and

preferred method to validate a simulation model (refer to

Figure 2). Assumptions, input values, output values,

workloads, configurations and network system behaviour

should all be compared with those observed in the real

network.

4. Propose Simulation Model Development

for Network Utilization and Traffic

Many different types of modeling and simulation

applications are used in various disciplines such as acquisi-

tion, analysis, education, entertainment, research and

training [24]. In the Figure 3, theoretical model is based

on a random distribution of service duration. “Request”

defines the way clients use the computer network to

request services, while, “Response” represents the way

COMPARING THE ACCURACY OF NETWORK UTILIZATION PERFORMANCE BETWEEN 341

REAL NETWORK AND SIMULATION MODEL FOR LOCAL AREA NETWORK (LAN)

Figure 1. Simulation model development methodology.

Figure 2. Simulation model verification and validation methodology.

clients receive services from the server. Simulation

model is divided as follows: 1) to study physical of real

heterogeneous network environment; 2) transform

physical of real heterogeneous network environment into

logical model; and 3) develop and implement the

heterogeneous simulation model.

4.1. Physical Model of Real Heterogeneous

Network Environment

Figure 3 shows the network heterogeneous environment

in real world. Then, it needs to transform from

heterogeneous environment in real world into logical

model. The logical model is the phase where

mathematical techniques are used to stimulate

heterogeneous environment.

4.2. Logical Model of Heterogeneous Network

Environment

Figure 4 depicts the open queuing network based on

Queuing theory (M/M/1) will use to develop logical

model of heterogeneous network environment for

network traffic utilization. Queuing theory is robust

enough to include many different combinations.

Parameters like bandwidth capacity, size of packet

services and number of clients are used to “characterize”

the application traffic.

342 M. N. ISMAIL ET AL.

Figure 3. Real heterogeneous network environment at main and branch campus.

Figure 4. Logical model of heterogeneous environment at main and branch campus.

4.3. Development of Heterogeneous Network

Environment Model

This section describes a simple analytical queuing and

little law theories that capture the performance

characteristics of network utilization and traffic opera-

tions. A link refers to a single connection between

routers and hosts. The link bandwidth is the rate at which

bits can be inserted into the medium. The faster

bandwidth the more bits can be placed on the medium in

a given time frame [25]. Table 1 shows the parameters

that have been used in the model development. In open

queuing network, the throughput of the heterogeneous

network environment is determined by the input rate in

the system. Table 2 summarizes all the parameters used

in the model.

COMPARING THE ACCURACY OF NETWORK UTILIZATION PERFORMANCE BETWEEN 343

REAL NETWORK AND SIMULATION MODEL FOR LOCAL AREA NETWORK (LAN)

The original Queuing theory is defined as an average

number of clients in the system (variable name is “N”) in

Equation (1). Equation (2) is defined as traffic intensity

use by clients in the system. Equation (1) and (2) are

derived based on logical model that has been designed to

meet requirements for heterogeneous network

environment. Logical model is derived and formulated

in a single service (homogeneous concept) as in

Equations (3), (4), (5), (6) and (7). Then, the logical

model is derived to the heterogeneous network

environment in Equations (8), (9), (10), (11), (12), (13)

and (14).

Figure 5 shows how the model has been formulated

(1)

(2)

from real network environment to simulation model. The

main valuable aspects of the simulation study is to

explain and understand real world phenomena that are

costly to perform in the laboratory or difficult to collect

in field experiments. A successful simulation model that

is able to provide a sufficiently credible solution that can

be used for prediction. Since it is not feasible to construct

a simulation model that represents all the details and

behaviors of the real network, some assumptions must be

made about the real network to construct a simulation

model. Therefore, a simulation model is an abstract

representation of real network environment.

Table 1. Notations for original queuing and little theories.

Model Parameters Meaning

N Average number of clients in the system

T Average time a client spends in the system (second)

Clients arrival rates

µ Service rate in second

1/µ Mean service times

ρ

Traffic intensity

Table 2. Notations for model development.

Model Parameters Meaning

N Average number of traffics on the network

T Average time of clients arrive on the network (second)

P (P1,P2,P3,...Pm) Various of services

P1 Client uses single service

Size of packet service request by client ( traffic)

Traffic response from server to clients

N

klient + server

Number of clients in second over single service

U

klient + server

Network traffic utilization usage based on number of clients in second over single

servic

e

C (C

LAN,

C

WAN

) Size of Bandwidth on LAN and WAN interface ports

U

hetergenes

Network traffic utilization usage for heterogeneous environment

Heter

klient + server

Number of clients and traffics over heterogeneous environment

Total size of packet services request by clients (traffic)

Jum_klient Number of clients

T

jum

Total number of clients access on the network in second

Client uses single service for accessing network server

(3)

(4)

N

klient + server

= (Minta + Balas) * (Jum_klient * T)

(5)

(6)

λ

T)*klient_Jum(*N

klient

µ

=

T)*klient_Jum*PN

(1serverklient

=

+

T)*klient_Jum()(N

serverklientserverklient

µ

⋅+=

+

Jumlah

µ

klient

µ

server

µ

T*N

λ

=

µλ

µ

λ

ρ

<<=;1

344 M. N. ISMAIL ET AL.

U

klient + server

= N

klient + server

/ C

(7)

Client uses various services for accessing network server in Heterogeneous Environment

T

Jum

= Jum_klient * T

(8)

(9)

(10)

(11)

(12)

where

P

1

+P

2

+P

3

+…+Pm = U

klient1

+U

klient2

+U

klient3

+…+U

klientm

=

jumlah

µ

(13)

U

hetergenes

= Heter

klient + server

/ C

(14)

Figure 5. Model and simulation development phases.

5. Verification and Validation of Simulation

Model with Real Network Experimental

Lab experiment is based on ideal network in which there

is no packet losses, no jitter in delays and network

bandwidth is sufficient for all requirements. While, real

experiment is based on real network and need to consider

as follows: 1) network bandwidth is limited and is not

enough for all application and users at the same time; 2)

delay due to the network overloads; and 3) packet losses.

5.1. Real Network Setup

This research used a network management application to

capture traffic between two networks link in real network

environment. Figure 6 shows the experimental setup of

real network used in our tests. The real network used

switch with Gigabit Ethernet ports, Router ports and

Fluke Optiview device can be configured to insert size of

packet services and number of clients to generate traffic

into the network interface (see Figure 7). By using

varying number of clients and size of packet services,

Fluke Optiview device is able to simulate network

utilization and traffic.

5.2. Real Network Experiment

This research has setup a real network environment of

network utilization measurement that generates

measurement data to analyze network performance at the

main campus. The real network is based on local area

network (LAN). The traffics will pump into LAN 100

Mbps (real network) to access network server (see Figure

7). Low bandwidth link affects the size of packet

services and number of clients’ access to the network

server. Therefore, network management application is

used to measure traffic and its network utilization

performance (see Figure 8). Five sets of experiments

were conducted with different scenarios (see Table 3 and

Jumm321serverklient

T*)P....PP(PHeter

+

=

+

Jumserverklient_mklient3klient2klient1serverklient

Τ*])...[(Heter

µ

+++++=

+

JumserverJumlahserverklient

T*)]()[(Heter

µ

+=

+

JumserverJumlahserverklient

T*)]()[(Heter

µ

+=

+

COMPARING THE ACCURACY OF NETWORK UTILIZATION PERFORMANCE BETWEEN 345

REAL NETWORK AND SIMULATION MODEL FOR LOCAL AREA NETWORK (LAN)

Table 4). Fluke Optiview device is able to generate

maximum traffic to 1518 bytes (12144 bits) only in the

real network (see Figure 7 and Figure 10). The same

input variables have been used in simulation model (see

Figure 9 and Table 4) to estimate our data that must be

closely resemble to real network environment (see Table

5). This research is concluded that base on our findings,

the simulation model is able to predict and estimate

network utilization usage for real network environment

(see Table 3 and Table 4).

5.3. Comparison of Real Network, Simulation

Model and Relative Error Rates

Figure 11 shows a comparison between simulation model,

real network and relative error rate using LAN 100 Mbps.

The result shows both scenarios use in simulation model

and real network are able to predict and measure net-

work traffic utilization. The simulation model provides

relatively accurate results when compared to the real

network over LAN 100 Mbps. Figure 11 also shows the

Figure 6. Experimental laboratory for real network environment setup.

Figure 7. Fluke Optiview engine setting for size of packet services and clients.

346 M. N. ISMAIL ET AL.

Figure 8. Real network experiment result capture by network management application (100 mbps).

Figure 9. Prediction of network traffic utilization over 100 mbps via simulation model.

comparison of relative error rates between simulation

model and real network environment. As a result, this

research shows that the simulation model can predict real

network experiments with minimum relative error rates.

Therefore, from the prediction and estimation result, this

simulation model can assist network administrator

COMPARING THE ACCURACY OF NETWORK UTILIZATION PERFORMANCE BETWEEN 347

REAL NETWORK AND SIMULATION MODEL FOR LOCAL AREA NETWORK (LAN)

Figure 10. Fluke Optiview device architecture.

Table 3. Network traffic utilization over 100 mbps for real network experiments.

Table 4. Network Traffic utilization over 100 mbps using simulation model.

Frame Size (Client +

Server); (Bytes)

Number of

Clients

(second)

Traffic in

Bytes/Second Traffic in

Bit/Second Utilization

(1-100%) Utilization

(0.1–1%)

512 262 134144 1073152 1.07 0.01073152

778 1633 1270474 10163792 10.1638 0.101638

831 149 123819 990552 0.99 0.0099

1042 3961 4127362 33018896 33.01 0.3301

1518 1726 2620068 20960544 20.96 0.2096

Table 5. Comparison of relative error rates between simulation model and real network environment.

Frame Size (Client +

Server); (Bytes)

Number of

Clients

(second)

Utilization

(Simulation Model)

(0.1–1%)

Utilization

(Real Network)

(0.1–1%)

Relative Error

Rates

778 812 0.01073152 0.0111 0.000119

831 393 0.101638 0.104 0.002362

1033 393 0.0099 0.0101 0.0002

1518 180 0.3301 0.337 0.0069

to plan, propose and design network topology more

systematic and efficiently for heterogeneous network

environment.

6. Conclusions and Future Work

This article has shown how an analytical queuing model

Frame Size

(Bytes)

Number of

Clients

(second)

Traffic in

Bytes/Second Traffic in

Bit/Second Utilization (1-

100%) Utilization

(0.1–1%)

512 262 134144 1073152 1.11 0.0111

778 1633 1270474 10163792 10.4 0.104

831 149 123819 990552 1.01 0.0101

1042 3961 4127362 33018896 33.7 0.337

1518 1726 2620068 20960544 21.2 0.212

348 M. N. ISMAIL ET AL.

Figure 11. Comparison of simulation model with real network using 100 mbps variable and relative error rate.

can be used to understand the behaviors of hetero-

geneous environment over LAN experiments. The most

apparent aspect is the utilization usage due to size of

bandwidth and number of clients. Our simulation model,

has demonstrated that it can measure accurately the

performance of heterogeneous services and technologies

to access network server. Through real network

experiments, the simulation model is verified and

validated for providing accurate performance infor-

mation for various services. The simulation-modeling

framework described in this study can be used to study

other variations, tunings, and similar new ideas for

various services and technologies. Network utilization

rate will directly affect the network performance. In

network management, by monitoring and analyzing

network utilization rate, network administrator can

monitor the performance of the network, thus to study

whether network is normal, optimal or overloaded.

Network utilization rate also plays an important role in

benchmark setting and network troubleshooting. Future

work is to develop a simulation model to analyze

bandwidth capacity requirement for various services and

technologies in heterogeneous environment.

7. References

[1] S. Kawasaki, Niwa, et al., “A study on formulation of the

ubiquitous cloud model”, Mobile Data Management,

2006, MDM 2006, 7th International Conference, pp. 148–

148, 2006.

[2] A. Tsalgatidou, G. Athanasopoulos, et al., “Developing

scientific workflows from heterogeneous services,” ACM

SIGMOD, Vol. 35 (2), pp. 22–28, 2006.

[3] Q. W. Liu, S. L. Zhou, et al., “Cross-layer modeling of

adaptive wireless links for QoS support in heterogeneous

wired-wireless networks,” Wireless Networks, Vol. 12

(4), Kluwer Academic Publishers, 2006.

[4] R. Puigjaner, “Performance modelling of computer

networks,” IFIP/ACM Latin America Networking

Conference 2003, La Paz, Bolivia. ACM 2003, October

3–5, 2003.

[5] S. Fredj, T. Bonald, A. Proutiere, G. Regnie, and J.

Roberts. “Statistical bandwidth sharing: A study of

congestion at flow level,” in Proceedings of ACM

SIGCOMM ’01, San Diego, CA, August 2001.

[6] G. J. Jin and B. L. Tierney, “System capability effects on

algorithms for network bandwidth measurement,”

Proceedings of the 3rd ACM SIGCOMM conference on

Internet measurement, pp. 27–38, 2003.

[7] J. Cong and B. E. Wolfinger, “A unified load generator

based on formal load specification and load

transformation,” Proceedings of the 1st international

conference on Performance evaluation methodologies and

tools, ACM International Conference Proceeding Series,

Pisa, Italy, Vol. 180 (53), 2006.

[8] S. B. Fredj, T. Bonald, and A. Proutiere, G. Regnie and J.

Roberts, “Statistical bandwidth sharing: A study of

congestion at flow level,” ACM SIGCOMM, August

2001.

[9] T. Bu and D. Towsley, “Fixed point approximation for

TCP behavior in an AQM network”, ACM SIGMETRICS,

June 2001.

[10] A. A. Kherani and A. Kumar, “Performance analysis of

TCP with nonpersistent sessions”, Workshop on

Modeling of Flow and Congestion Control, INRIA, Ecole

Normale Superieure, Paris, September 4–6, 2000.

[11] Q. G. Zhao, X. M. Fang, Q. Z. Li, and Z. Y. He, “WNN-

based NGN traffic prediction,” ISADS 2005, Proceedings

COMPARING THE ACCURACY OF NETWORK UTILIZATION PERFORMANCE BETWEEN 349

REAL NETWORK AND SIMULATION MODEL FOR LOCAL AREA NETWORK (LAN)

Autonomous Decentralized Systems, pp. 230–234, 2005.

[12] C. Barakat, P. Thiran, G. Iannaccone, C. Diot, and P.

Owezarski. “Modeling Internet backbone traffic at the

flow level”. IEEE Transactions on Signal Processings

(Special Issue on Networking), August 2003.

[13] T. Bu and D. Towsley, “Fixed point approximation for

TCP behavior in an AQM network,” in Proceedings of

ACM SIGMETRICS ’01, San Diego, CA, June 2001.

[14] P. Podhradsky, “Migration scenarios and convergence

processes towards NGN (present state and future

trends),” Electronics in Marine Proceedings Elmar. 46th

International Symposium, pp. 39–46, 2004.

[15] G. Nogueira, “Measurement based validation of GPRS/

EDGE analytical models”.

[16] J. Heidemann, K. Mills, and S. Kumar, “Expanding

confidence in network simulation,” IEEE computer, Vol.

15 (5), pp. 58–63, 2001.

[17] C. Barakat, P. Thiran, G. Iannaccone, C. Diot, and P.

Owezarski, “A flow-based model for Internet backbone

traffic,” Proceedings of the 2nd ACM SIGCOMM

Workshop on Internet measurement, Marseille, France,

pp. 35–47, 2002.

[18] J. kontio and R. Conradi, “Evaluating the performance of

a web site via queuing theory,” Software Quality - ECSQ

2002: 7th International Conference, Helsinki, Finland, pp.

63–72, Springer-Verlag Berlin Heideberg, June 9–13,

2002.

[19] K. H. Lee, K. O. Lee, K. C. Park, J. O. Lee, and Y. H.

Bang, “Architecture to be deployed on strategies of next-

generation networks,” Communications ICC ’03, The

IEEE, Vol. 2, pp. 819–822, 2003.

[20] J. Q. Li, Z. H. Sun, 2004, “Internet/Web technology in

higher education in China,” Advanced Learning

Technologies, Proceedings IEEE International

Conference, pp. 993–997, 2004.

[21] X. X. Jiang, F. C. Yang, and H. Zou, “A novel

architecture to customer service management for the

NGN,” ICCT 2003. International Conference, 1: 123–126,

2003.

[22] B. Thai, R. Wan, A. Seneviratne, and T. Rakotoarivelo,

“Integrated personal mobility architecture: A complete

personal mobility solution,” Kluwer Academic Publishers,

2003.

[23] J. Curtis and T. McGregor, “Review of bandwidth

estimation techniques,” in Proceedings New Zealand

Computer Science Research Students’ Conference, Vol. 8,

New Zealand, April 2001.

[24] D. R. Gerhan and S. M. Mutula, “Bandwidth bottlenecks

at the University of Botswana,” Published by Emerald

Group, Vol. 23 (1), pp. 102–117, 2005.

[25] J. Sommers and P. Barford, “Self-configuring network

traffic generation,” in Proceedings of ACM Internet

Measurement Conference, October 2004.

Paper Menu >>

Journal Menu >>