See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228740465

Performance analysis of network operating systems in local area networks

Article · January 2008

CITATIONS READS

3 2,674

4 authors, including:

Shaneel Narayan Samad Kolahi

UNITEC Institute of Technology UNITEC Institute of Technology
147 PUBLICATIONS   2,585 CITATIONS    56 PUBLICATIONS   379 CITATIONS   

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Samad Kolahi on 29 May 2014.

The user has requested enhancement of the downloaded file.

 2nd WSEAS Int. Conf on COMPUTER ENGINEERING and APPLICATIONS (CEA'08) Acapulco, Mexico, January 25-27, 2008

Performance Analysis of
Network Operating Systems in Local Area Networks
SHANEEL NARAYAN, SAMAD KOLAHI, RICK WAIARIKI and MADELEINE REID
School of Computing and Information Technology
UNITEC New Zealand
Carrington Road, Auckland
NEW ZEALAND
snarayan@unitec.ac.nz http://www.unitec.ac.nz

Abstract: - In this paper, in a laboratory environment the performance of four different operating systems
(Windows NT4, Windows 2000, Windows 2003, and Linux Fedora) are compared. The performance
parameters measured are bandwidth and network delay. Linux Fedora provided the highest bandwidth for a
file server at 17.1Mbps, Windows 2000 was fastest for a FTP (83.2Mbps), and Windows 2003 and Windows
2000 gave the highest bandwidth for a web server (4.3Mbps and 4.5Mbps respectively).

Key-Words: - Operating systems, performance analysis, Windows NT4, Windows 2000, Windows 2003, Linux.

1 Introduction approach is novel in the sense that it measures

With the increased uses of computer networks, the
performance analysis of networks is becoming critical.
Businesses need to transfer more data at the minimum
amount of time in their LANs and WANs. At the heart
of a client-server environment are network operating
systems. These modern operating systems contain
significantly more embedded features than what their
predecessors had, thus it is necessary to know how they
perform with these increased functionality. However,
comparison of these operating systems is not an easy
task as different operating systems have different
features and functionality.
To the authors’ knowledge, there is little work done to
compare the performance of network operating systems
in terms of delay and bandwidth using a test bed in a
laboratory. Vendor sponsored research is available,
however results are biased. For example, IT Week Labs
indicate that Linux Samba is 2.5 times faster then
Windows Server 2003 [1], while Microsoft sponsored
research carried out by Veritest found Windows Server
2003 outperforms the counterpart in common aspects of
client-server activities [2]. Performance of web related
activities have been undertaken by some researchers [3-
6] looking at specific aspects of just web server.
Evaluating how subsystem interaction with operating
system affects performance has also been researched [7-
9]. Sprunt [10] has studied how application and
operating system perform on different computer
processors. Within the scope of related work, our
throughput and delay on a test bed set-up.
The rest of the paper is organised as follows:
Experimental set-up and methodology is described in
Section 2. Section 3 reports the experimental results and
discusses. Finally the paper is concluded in Section 4,
followed by references.
2 Experimental Setup
In a laboratory setting (Figure 1), a LAN network with
TCP/IP protocol is set-up and the operating systems
were changed in order to compare the results. All
computers and servers are connected in star topology to
a 10/100 Ethernet switch with 100Mbps UTP links. The
setting included a file server, FTP server, Web server,
DHCP and DNS server, Domain Controller and a
workstation. The hardware specifications for all
computers on this network are IBM 8305-HAD machine
type, Intel Celeron 2GHz processor, 256MB RAM, and
Intel Pro 100S network adaptor. This being a
comparative performance measurement study, it is
imperative that the test bed infrastructure remains
constant for all the operating systems
.
On the test bed infrastructure, each operating system
(Windows NT4, Windows 2000, Windows 2003 and
Linux Fedora) is installed one at a time. For each, latest
patches and services packs were installed in accordance

ISSN: 1790-5117
Page 186 ISBN: 978-960-6766-33-6
 2nd WSEAS Int. Conf on COMPUTER ENGINEERING and APPLICATIONS (CEA'08) Acapulco, Mexico, January 25-27, 2008
with the vendors’ instructions and no optimisation of
any sort was performed on the setup. Each operating
system was also matched with appropriate services and
applications (for example, Windows 2003 was matched
Figure 1. Illustration of the laboratory setup
We used NetBench, WebBench and FTPBench as the
primary tools for measuring performance parameters for
the network. NetBench measures performance
parameters of file servers as it fulfils requests from
client computers. This program has been designed for
client/server environment, thus it simulates file transfers
of word processed and spreadsheet documents from a
file server to workstations. WebBench measures similar
parameters by generating requests from client
workstations to web servers for combination of static
files and dynamic executables that run in order to
produce data that the server returns to the client. Web
browser is simulated on the workstations which make
requests to the server. The client records information
such as how long the server took and how much data is
returned. FTPBench works in a similar way to
Webbench but uses the GET command to transfer files
of various sizes from an FTP server to the workstation.
All three programmes stress test the test bed by
simulating up to 60 concurrent connections for each
operating system.
3 Results and Discussion
For various operating systems (Windows NT4,
Windows 2000, Windows 2003, Linux Fedora) the
bandwidth is reported in Table 1. As a file server, the
ISSN: 1790-5117
Page 187
with IIS version 6 and Linux Fedora with Apache
version 2.0.49). DNS (for name resolution) and DHCP
(for automation of TCP/IP parameters allocation) were
also implemented for each operating system.
bandwidth recorded for all operating systems were
between 16.2Mbps and 17.1Mbps. While Linux Fedora
marginally performed better than other operating
systems, Windows 2000, Windows 2003 and Windows
NT4 had close bandwidth. However, as a web server,
Windows 2000 and Windows 2003 performed better
than the rest with 4.5Mbps while Windows NT4 had the
worst performance with only 1.2Mbps. Windows NT4
did not perform well here as it does not have advanced
caching that the other operating systems use for web
servers. FTP performances of operating systems were
Windows 2000 with 83.2Mbps (the best), Windows NT4
with 77.2Mbps and Windows 2003 only at 55.1Mbps
(the worst). The FTP test for Linux was not conclusive.
Windows 2003 did not perform well with FTP possibly
because of increased functionality and features that can
sometimes slow the system. The results further show
that the actual bandwidth is less than the capacity of the
link (100Mbps) and that operating system cannot take
full advantage of the link bandwidth. FTP had the
highest link usage (77-83Mbps) as FTP is especially
made for file transfer while web server had the least link
capacity usage of between (1.3 to 3.7Mbps).
File Web FTP
Server Server (Mbps)
(Mbps) (Mbps)
Windows 16.35 4.5 83.2
2000
ISBN: 978-960-6766-33-6
 2nd WSEAS Int. Conf on COMPUTER ENGINEERING and APPLICATIONS (CEA'08) Acapulco, Mexico, January 25-27, 2008

Windows 16.20 4.3 55.1
2003
Windows 16.60 1.2 77.2
NT4
Linux 17.1 3.7
(Fedora)
Table 1: Comparative results of bandwidth versus
operating systems for the four operating systems
compared.
The network delay was also recorded for various
operating systems. For file transfer, delay varied from
0.57msec to 0.45msec with Linux providing the least
delay as it has the fastest bandwidth for data transfer.
As web server, WindowsNT4 creates much delay of
15msec due to low bandwidth of only 1.2Mbps while the
network delay of other web servers were between 3.8-
4.0msec.
4 Conclusion
The performance of four operating systems (Windows
NT4, Windows 2000, Windows 2003, and Linux
Fedora) were compared being used as a file server, web
server, or FTP server. The results indicated that Linux
Fedora provided the best bandwidth as a file server
(17.1Mbps) and Windows 2000 for file transfer at 83.2
Mbps. Windows 2003 and Windows 2000 at 4.3Mbps
and 4.5Mbps respectively were the fastest as a web
server while Windows NT4 was the slowest due to
caching issues.
[4] A. Shukla, L. Li, A. Subramanian, A. S. Ward, and
T. Brecht, "Evaluating the performance of user-space
and kernel-space web servers," in Proceedings of the
2004 conference of the Centre for Advanced Studies
on Collaborative research. Markham, Ontario,
- Canada: IBM Press, 2004.
[5] E. B. Nahum, T.; Kandlur, D.D, "Performance issues
in WWW servers," IEEE/ACM Transactions on
Networking, vol. Volume 10, pp. 2-11, 2002.
[6] V. Cardellini, E. Casalicchio, M. Colajanni, and P.
Yu, S. , "The state of the art in locally distributed
Web-server systems," ACM Comput. Surv., vol. 34,
pp. 263-311, 2002.
[7] M. Swift, M. , B. Bershad, N. , and H. Levy, M. ,
"Improving the reliability of commodity operating
systems," ACM Trans. Comput. Syst., vol. 23, pp.
77-110, 2005.
[8] E. Billard, A. , "Operating system scenarios as Use
Case Maps," in Proceedings of the 4th international
workshop on Software and performance. Redwood
Shores, California: ACM Press, 2004.
[9] J. Russell, T. and M. Jacome, F. , "Architecture-
level performance evaluation of component-based
embedded systems," in Proceedings of the 40th
conference on Design automation. Anaheim, CA,
USA: ACM Press, 2003.
[10] B. Sprunt, "The basics of performance-
monitoring hardware," Micro, IEEE, vol. 22, pp. 64 -
71 2002.

References:
[1] R. Howort, "Samba 3 extends lead over Win 2003,"
IT Week:
http://www.itweek.co.uk/itweek/news/2085218/samb
a-extends-lead-win-2003, 2003.
[2] Veritest, "Microsoft Windows Server 2003 Standard
Edition vs. Samba 3.0 and Red Hat Enterprise Linux
ES 3.0 File Server Performance Comparison,"
http://www.lionbridge.com/NR/rdonlyres/6BE8466A
-738D-4952-B61C-
4CB4070DC368/0/ms_samba.pdf, 2004.
[3] C. Coarfa, P. Druschel, and D. Wallach,
"Performance analysis of TLS Web servers," ACM
Trans. Comput. Syst., vol. 24, pp. 39-69, 2006.

ISSN: 1790-5117
Page 188 ISBN: 978-960-6766-33-6

View publication stats