IOP Conference Series: Materials Science and Engineering

PAPER • OPEN ACCESS

Network Topology Detection System

To cite this article: G. B. Abdulhussein et al 2018 IOP Conf. Ser.: Mater. Sci. Eng. 433 012085

View the article online for updates and enhancements.

This content was downloaded from IP address 181.215.64.245 on 01/12/2018 at 00:48

2nd International Conference on Engineering Sciences IOP Publishing
IOP Conf. Series: Materials Science and Engineering 433 (2018) 012085 doi:10.1088/1757-899X/433/1/012085
1234567890‘’“”

Network Topology Detection System

*
G. B. Abdulhussein1, T. Kanakis2, A. A. R. Altahir3
1
Computer Network Engineering Dept. - Faculty of Arts, Science & Technology – University of
Northampton, Northampton, UK
2
Faculty of Arts, Science & Technology – University of Northampton, Northampton, UK
3
Electrical Engineering Department, Faculty of Engineering, University of Kerbala

Email: [email protected]

Abstract. Detecting network topology is essential to cope with any existing or future network
changes such as the loss of network devices (routers, switches, or hubs), interfaces going down,
or cable issues, any if which might change packets’ paths, interfere with the network structure,
interface status, or neighbour status, or disconnect and reconnect devices. This paper develops a
low-cost network detection system that is responsible for discovering network topology by
identifying each network device's neighbours to allowed the network administrators to know
where each device and its neighbours are located, maintaining these devices’ information in a
local database to be used for visualisation purposes to demonstrate relevant changes inside the
devices such as IP address, interfaces, and neighbours by using a configured hardware device.
This system can be implemented in any large to medium size of network, measured depending
on the number of users that join such a network. In addition, it can be used to remotely fix any
minor issues that are detected which may later affect the entire network. The experimental
results substantiate the fact that the system resolves the high cost of hardware experienced by
previous work in this field.

1. Introduction
Discovering network topology is very significant, as increasing companies’ network structure makes
them become bigger and more complex, and thus inserting network data into a database manually has
become difficult. As well as dealing with any existing and future adjustments in network structure,
detecting and resolving network failures takes time to achieve [1] and [2]. There are a number of
hardware devices that can perform this function. However, the hardware device price is generally high
compared with the network topology size, and the administrators are required to insert the network
topology manually into these systems. The current system thus focusses on allowing the network
managers to display more information on their networks by discovering connected devices using
configured hardware; experimental work suggests that this depends on the available devices on the
network, along with their features [3].
This paper is organised as follows: Section 2 examines previous work that had been produced in this
field, while section 3 deals with the experimental work of this paper, including choosing the hardware
and the programming language before implementing the proposed system on the chosen hardware
device. Section 4 discusses the experimental results of the system, and includes some screenshots.
Finally, section 5 draws some conclusions about the implementation of the system and the results, as
well as offering future recommendations.

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
2nd International Conference on Engineering Sciences IOP Publishing
IOP Conf. Series: Materials Science and Engineering 433 (2018) 012085 doi:10.1088/1757-899X/433/1/012085
1234567890‘’“”

2. Literature review
Some researchers have acknowledged previous art in the tools listed in their patent that embody the
observation and control of network performance. Alongside the strengths of their tool, there are a
number of weaknesses about which they do not offer sufficient information, though these are readily
distinguishable as the source of the problem in the network [4].
A current system that applies the same approach [5] illustrates the progress of the Customer Operation
Service Management Interface Control system (COSMIC) in order to observe the private huge
telecommunication used by the AT&T corporation network control centre. This showcased network
health status monitoring using individual computers. Its major aptitude is displaying simultaneous
network management outcomes. However, the architecture of the hardware is relatively expensive.
LAN status monitoring was investigated in [6]; however, this work merely considered token ring, FDDI,
and Ethernet networks. In [7], an effective and automatic network monitoring system that continuously
monitors network switches to inform the network managers by SMS or Email when a switch is lost by
using smart communication Request Tracker (RT) and Nagios software available in a Linux
environment was investigated; more negatively, this system generates tickets for each failure and only
takes the issue’s ticket into consideration when resolving it.
The authors in [8] aimed to monitor network architecture by installing a proxy system in all the devices
of the network which acts as a requisite for interconnecting with the central management unit remotely.
In each device, the data are gathered and sent to the central management unit over the internet as a
message. The strengths of the patent are the discovery of failed devices in order to report these to the
administrators. However, this system must be installed on each device and each device must be able to
connect remotely with the central management unit, which means that whenever a new device joins the
network, a new agent system must be installed.
The patent authors’ in [9] indicated that their network management station used a tool and technique
with the purpose of discovering and monitoring all the devices on the network simultaneously to display
the outcomes in a desktop-based application with all associated information.
Based on the patent in [10], those researchers concentrated on a method to detect the topology of the
network. The authors built the network from their own design, which was provided before applying their
technique in order to create a comparison. Using the invented method will lead to detection of an actually
constructed network topology, and matching the designed topology with the detected topology is the
aim. This requires the designed topology to be identified for the matching step, and as such, this is
restricted to applications on small networks.
Based on the patent in [11], these inventors focused on a technique to detect the topology of the network
to produce a drawing that could be useful for both network managers and designers. The topology
producer collects information that from both the ARP (Address Resolution Protocol) cache and MAC
(Media Access Control) caches of the network devices. It depends on the gathered data to perform the
topology detection stage and uncovers links or unknown domains in the network devices. The
association data may be used to define the cables in the topology along with the type of network devices
found.
In this paper, the focus is on the selection of a suitable hardware device to discover the construction of
a network in order to access all the network devices, in order to create a main copy of the network
topology which will be modified as network devices change. Moreover, a low-cost constraint will be
applied for purchasing and implementing when compared with current hardware. This system will be
developed to be harmonious with any medium to large network where the size of the network is
measured in terms of the number of users that connect to the same network structure, including
universities, companies, and other business sector networks.

3. Experimental Work

3.1. Hardware selection

A number of hardware devices could be used to implement the proposed system. Recently, the most
popular hardware devices that have been used are Raspberry Pi and Arduino, due to their distinctive

2
2nd International Conference on Engineering Sciences IOP Publishing
IOP Conf. Series: Materials Science and Engineering 433 (2018) 012085 doi:10.1088/1757-899X/433/1/012085
1234567890‘’“”

features. Arduino does not have an operating system which can be run under Arduino IDE as a platform,
however, so only one program could be run at a time, and so on. Table 1 shows the comparative technical
features of Raspberry Pi and Arduino.
Table 1: Raspberry Pi and Arduino features

Feature On Raspberry Pi On Arduino

Backwards pin 40 pin GPIO connector Programmed with FTDI breakout
compatibility boards
Processor Choice a Broadcom BCM2837 SoC with 1.2 GHz ATmega238P
64-bit quad-core ARM Cortex A53
processor
Memory 1 GB RAM 32 KB
Operating System Customised Debian Linux: Raspbian None
price Between $5 and $35 22.57$
Based on Single board computers Microcontroller based development
boards
Based on these features, Raspberry Pi was selected for the experimental work. The price of a Raspberry
Pi currently starts at around $5 (raspberry pi zero), rising to $35 for a board with a power supply without
a case, HDMI cable, or RAM. Additional components will cause the price to rise. The cost of the basic
Arduino board is higher, and as the Raspberry Pi is a micro-computer that runs under the Raspbian
operating system, it can also be used as a separate system if provided with a mouse, keyboard, and
monitor without requiring an additional operating system as in the case of Arduino. A number of ports
are available to connect such items, and a number of programs can be run at the same time on Raspberry
Pi, while only one program can be executed at a time with Arduino. Further, Raspberry Pi supports
802.11n Wireless LAN (Local Area Network) and Bluetooth and offer good open source documentation.
[12-15, 3]

3.2. Programming language selection

Raspberry Pi supports a number of programming languages that could be used to detect network
topology such as Python and Java, which are preinstalled with the operating system [14]. As Python is
the most commonly used programming language, it can be used as an interpreter for another program,
while Java requires a compiler that runs from the processor. It is also object-oriented, open source, and
fast as compared with Java, which is slow in processing; Python was thus used to implement the
proposed system [13, 16].

3.3. The experimental work of proposed system

A network detection system is used to access network devices such as routers and switches remotely via
the connected network device, which is the router that connects to the entire network on one side and to
the outside world on the other; to do this, the Raspberry Pi is connected using an Ethernet cable [3].
Figure 1 illustrates the connection between the case study network and the Raspberry Pi. Telnet [17] is
the protocol used on the internet or Local Area Network (LAN) in order to offer interactive command
line communication capability for remotely accessed computers [18] or network equipment [19]. Thus,
this is used to obtain access to the connected network device via the Raspberry Pi, and through this
connection, the whole network topology can be accessed, along with data such as IP addresses,
interfaces, neighbours, and hostnames. This is done by reading the first accessed router's data in order
to find its neighbour(s), and then generating a loop that goes through these one by one remotely, using
Telnet commands to find other neighbours and data by sending different network commands to the
accessed network device. Each time a new neighbour is found that has not been accessed previously, it
is accessed in turn until the whole network's devices are stored in a database (DB). The program will
execute repeatedly to find any network changes.fi

3
2nd International Conference on Engineering Sciences IOP Publishing
IOP Conf. Series: Materials Science and Engineering 433 (2018) 012085 doi:10.1088/1757-899X/433/1/012085
1234567890‘’“”

Figure 1: Case study network connected to Raspberry Pi

The initial connected network device's (first router) IP address must be known, and the network device
credentials that allow remote connection through Telnet must be provided to the system in order to
assess all devices. An overview of the proposed system is shown in figure 2.

Figure 2: Network Detection System Overview

The main function of this system is to find the devices on the network through the main router connected
to the Raspberry Pi; depending on the information from this device and its neighbours, any new device
can be accessed remotely, which leads to the detection of other network devices such as routers, and
switches, until the complete network topology is known. After that, the collected data are stored in a DB
[20] which continues to close the connection with the current device to access another device until the
whole network discovered; this process is then repeated every ten seconds after the whole network is
determined [3]. The time for re-executing the program depends on the size of the network, which means
that whenever the network increases, the time required increases as well. This experimental work was
part of a master's dissertation [3], which used the DB in its later stages to create a website for
visualisation, network health status tests, and real-time updates.

4. Experimental Results
The network topology was discovered using a Raspberry Pi in a format that connects to any medium to
large size network using an Ethernet cable, which is authorised by the system’s user to find all the
attached devices. The data is collected via communication from the source to the destination. Whenever
the hardware device accesses a network device, a new file is created or overwritten on the same file if it
already exists in order to store and extract the network device data obtained through sending different
network commands, as shown in figure 3; this includes the name of the device, IP addresses, interface
status, and neighbours.

4
2nd International Conference on Engineering Sciences IOP Publishing
IOP Conf. Series: Materials Science and Engineering 433 (2018) 012085 doi:10.1088/1757-899X/433/1/012085
1234567890‘’“”

Figure 3: Network Device File Content (Router)

All the network topology data is stored in a local database that can be managed by the network
administrator; part of this is listed in Table 2. This step has not been provided in previously published
work, where databases were inserted manually into the system or only notifications that a device has a
problem are provided [6]. High-cost hardware and software [5] solutions also cannot compare with this
information gathering approach.
Table 2: Sample Database Results
Device_name IP Interface int_status int_portocol network_add connect_to connect_to_int

R_edge 192.168.1.1 FastEthernet0/0 up up 192.168.1.0/24 Switch 192.168.1.2

R_edge 192.168.10.1 FastEthernet0/1 up up 192.168.10.0/24 NULL NULL

R_edge unassigned Serial0/0/0 administrativelydown down None NULL NULL

Active_core 192.168.1.3 FastEthernet0/0 up up 192.168.1.0/24 Switch 192.168.1.2

Active_core 192.168.2.3 FastEthernet0/1 up up 192.168.2.0/24 Switch 192.168.2.2

S1_edge unassigned FastEthernet0/1 up up None Router 192.168.1.1

Network detection time plus ten seconds is the time used for repeating the same process, though this
could be changed depending on the network administrators' detection of network changes, This includes
the storage of the network topology information in a database, which reads the contents of the file in
detail. The repeating process time thus relies on the number of network devices. The experimental results
demonstrate that the system can solve the high-cost issue of buying and implementing systems to allow
any network to detect its network topology. Thus, the achieved outcomes of checking the system
promote the significance of obtaining true network topology with a low-cost network discovery system
when compared with the AT&T company hardware used to detect and monitor that company's network
[5, 12]. However, the time of detection process remains an issue; this could be minimised by accessing
the network devices simultaneously using a number of Telnet sessions.

5. Conclusion
A network detection system has been proposed to minimise the system costs, which matches network
administrators’ requirements. The results of testing the system support the relevance of obtaining
accurate and filtered network device data at lower cost. In this paper, the discovery process of the devices
connected to a network was carried out using a Raspberry Pi as a configured hardware device. Taking
into consideration the fact that the core router might not have wireless signal, an Ethernet cable is
essential to form a connection between the Raspberry Pi and the network device that connects the overall
network with the outside world. The device data is then stored in a new or existing file using the name
of the network device that has been accessed. After the process of discovering the first router, storing
the extracted data from the network device is performed subsequently. When the first router's data is
stored in the DB, its first neighbour is then accessed remotely and its device data stored, and so on until
the whole network topology is discovered. As a result, this system solves the issue of the high cost of

5
2nd International Conference on Engineering Sciences IOP Publishing
IOP Conf. Series: Materials Science and Engineering 433 (2018) 012085 doi:10.1088/1757-899X/433/1/012085
1234567890‘’“”

existing systems that only the massive networks can effectively use and offers a system that can be used
by any network from medium to large size [5]. The time for discovering a network's topology depends
on its size, which means that every time the network grows in size, the time is increased. To decrease
this time, a number of Telnet sessions could be run at the same time, as future recommendations to
improve the system.
Future recommendations to improve the system include the fact that the database could be stored in the
Cloud to offer a recovery method for the network topology when the company replaces hardware, and
more information on the network could be used in order to provide network administrators with fuller
data on each device.

References

[1] Oliveira O M F 2017 Network topology discovery University of Aveiro Department of

Mathematics, Thesis.
[2] JENSEN R 2007 Network discovery system U.S. Patent 20080155386A1.
[3] Abdulhussein G B A 2017 Network Health Status Monitoring University of Northampton, MSc
Dissertation (UK: Northampton).
[4] Fletcher R A, Banthia P C and Svensson A 2001 Integrated Computer System and Network
Performance Monitoring U.S. Patent 6,269,401 B1.
[5] Chang L, Chan W L, Chang J, Ting P and Netrakanti M 1989 A network status monitoring
system using personal computer IEEE Global Telecommunications Conference 1989 and
Exhibition Communications Technology for the 1990s and Beyond vol. 1 pp. 201-206.
[6] Choi T and Tang A 1995 Enterprise network management: LAN status monitoring
In Communications, 1995. ICC'95 Seattle, 'Gateway to Globalization', 1995 IEEE
International Conference on Vol. 3, pp. 1448-1452.
[7] Khan R, Khan S U, Zaheer R, and Babar M I 2013 An Efficient Network Monitoring and
Management System International Journal of Information and Electronics Engineering Vol.
3(1), pp. 122-126.
[8] Rackus P, Carter C, Fauteux J and Gilbert A 2005 Multi-network monitoring architecture U.S.
Patent 20050235058 A1.
[9] Chin J, Tran H and Leong L 2002 Method and apparatus for displaying health status of network
devices Patent US 6,456,306 B1.
[10] Anderson K W, Weichhold P and Simon C S 2005 Network Topology Discovery US Patent
10/907,929.
[11] Alexander W M, Shah D and Gibson W S 2013 Network Topology Discovery US Patent
13/984,976.
[12] Atwell C 2015 Top 10 Single Board Computers From $9 to $569 EETimes.com.
[13] Dennis A K 2016 Raspberry Pi Computer Architecture Essentials Packt Publishing Ltd
(Birmingham, B3 2PB ,UK).
[14] Gay W 2014 Raspberry Pi Hardware Reference Apress (New York, NY 10013).
[15] Shah S 2015 Learning Raspberry Pi Packt Publishing Ltd (Birmingham, B3 2PB, UK).
[16] Donat W 2014 Learn Raspberry Pi Programming with Python Apress (New York, NY 10013).
[17] Postel J and Reynolds J 1983 Telnet protocol specification [Online]. Available:
https://tools.ietf.org/html/rfc854. [Accessed 22 Nov 2017].
[18] MORRIS H C and CALABRESE J S 2002 Detecting an active network node using a login
attempt U.S. Patent 6,347,339.
[19] LEONG L and LEE F 2001 Method and system for monitoring remote routers in networks for
available protocols and providing a graphical representation of information received from the
routers U.S. Patent 6,269,398.
[20] MATHENY D, TAYLOR D and WINTERTON R 2001 System and method for collecting,
aggregating, and coalescing network discovery data U.S. Patent 09/846,521.