Professional Documents
Culture Documents
Convergence of Routing Protocols in Real and Simulated Environments
Convergence of Routing Protocols in Real and Simulated Environments
Abstract — In this paper, the comparative analysis of note that the most important information for routing
dynamic routing protocols EIGRP and OSPF is presented. protocols is information about state of local network. The
The objective of analysis was to determine which routing purpose of routing protocols is to determine the best path
protocol is converging faster. For testing, three approaches to destination network. This decision is done based on
were used, namely simulation using OPNET simulator,
metric information. The time needed to exchange routing
emulation using GNS3 emulator and testing using real
devices. Network topology that was used for testing purpose information and calculate the best path to the target
provide three possible paths between source and destination network is a time to achieve convergence. Authors in [2]
device. Generating large amount of echo request messages point out that the most common reasons for change in
and observing echo reply message before and after link computer network is a link failure, a metric change, and a
failure provide time needed to achieve convergence in network device failure. All these changes will cause
network. Comparing the results obtained using real devices calculations in dynamic routing protocols.
and simulator / emulator showed some shortcomings of
OPNET and GNS. From the test results, it can be said that A. Enhanced Interior Gateway Routing Protocol
the dynamic routing protocol EIGRP is converging faster. EIGRP is an advanced distance vector routing protocol
This protocol is therefore more suitable for implementation that is used in LAN environment to share routing
in production infrastructures.
information with neighboring routers in order to route
Keywords— dynamic routing protocols; EIGRP; OSPF; traffic within autonomous system. In publication [3] is
OPNET; GNS said that it is only protocol that ensures loop free
topology at any time since it is implemented. For
I. INTRODUCTION calculate paths to the remote networks is used Diffusing
Update Algorithm.
Usually, in a business environment, are used services
and software systems, which are sensitive to the delay of
the data being transmitted. It is therefore appropriate to
implement a dynamic routing protocol that will provide
the shortest time to recover from the link failure.
During the initial design and also during the change of
the already functioning infrastructure we have to test the
planned implementation or change in the network at first.
We are able to test routing protocols using computer
simulation or emulation programs. Selecting the routing
protocol is one of the most important questions when
designing a computer network. All protocols ensure
Figure 1. Finite state machine Dual algorithm [1]
convergence, but after different time interval has elapsed.
The speed of convergence is possible to determine based
In Fig. 1 is shown finite state machine of DUAL
on the theoretical knowledge of individual protocols, but
algorithm. The network is in a passive state when it is
many real situations can affect standard behavior.
stable. If change occurs, we lose information about
The aim of this work is to verify and compare duration
reaching the network, so we go to an active state in order
of convergence dynamic routing protocols on a selected
to trying to get information to reach the lost network
scenario, both in simulated environments and on real
through other neighbors.
facilities under laboratory conditions. Subsequently, the
The advantage of this protocol is that change in the
individual results will be compared and analyzed.
network cause only calculation on routers that is
II. DYNAMIC ROUTING PROTOCOLS influenced by the change. Research conducted by [4]
highlights that protocol EIGRP is suitable for big
In the following section the most commonly used complex networks because provide fast convergence and
dynamic routing protocols in local area network is consumes less bandwidth than other routing protocols.
introduced. Routing protocols EIGRP and OSPF are the
best choice for implementation in LAN. In [1] authors
425
Authorized licensed use limited to: ULB Darmstadt. Downloaded on December 19,2022 at 21:32:52 UTC from IEEE Xplore. Restrictions apply.
ICETA 2018 • 16th IEEE International Conference on Emerging eLearning Technologies and Applications • November 15-16, 2018, Starý Smokovec, The High Tatras, Slovakia
B. Open Shortest Path First enable it to be used to simulate both simple and complex
OSPF is a link state routing protocol operating within topologies and models.
autonomous system. All routers in AS have information It is suitable not only for computer network experts but
about all networks and connections between network even for beginners. In the program itself, it is possible to
devices in topology. In case of change in network work at different levels. From the lowest level where we
topology, routers that detect change have to advertise this can develop protocols and modules using C++, through
event to all routers in area. For calculating paths to the the creation of finite state machines, to their use, and
remote networks is used Djikstra’s shortest path first modeling computer network.
algorithm. In [5], author note that the most important part In publication [7] is said that OPNET Modeler uses
of OSPF process is synchronization of topology discrete event-driven simulation. The simulation time is
databases. The synchronization occurs during initial updated after performing the next event. This approach
establishment of adjacency or in case of topology change. leads to a more efficient use of system resources. DES
creates a detailed packet model, resulting in achieve
accurate simulation. This property is a disadvantage
because it increases the simulation time in case of
topology that includes several computing devices. To
solve this problem it is used in simulations not only DES
but also is used mathematical analysis. This combined
model is called hybrid simulation. In this type of
simulation DES computes interesting data and an
analytical method are used for simulation of all data
running in the background. Hybrid simulation helps
achieve the accuracy of the monitored data and save time
and resources.
B. Graphical Network Simulator
GNS3 is a free available graphical network emulator.
Figure 2. FSM establishing adjacency [1] Following the [8] GNS3 is most commonly used by
networking professionals to emulate network devices
In Fig. 2 is shown the process of establishing adjacency such as Cisco and Juniper, and also emulates virtual PCs
between connected routers. As we can see, there are eight and in a limited extent, ethernet switches. The advantage
states from which exchange and loading take the most of this software is possibility of use on various operating
time. Number of routers in topology influence the time systems (Windows, Linux, OS X). GNS can be also
needed for achieve convergention. connected with VirtualBox, so it is possible to simulate
III. SIMULATION AND EMULATION SOFTWARE real topologies using multiple endpoints with different
operating systems. Directly in this system is a Wireshark
Simulators and emulators can help network that is a network analyzer, which allows perfect control
administrators to test configuration before implemented over network communication. Another advantage is the
into production. This type of software reduce money ability to connect the GNS environment with the physical
needed for buying real devices and also reduce time to network interface and thus integrate it with a real-world
gain result of tested topology. Simulators can simulate network.
complex scenarios consisting of several devices in a short
time. It is possible to configure lots of application and IV. TESTING CONVERGENCE TIME OF DYNAMIC
services into topology and then observe the impact of the ROUTING PROTOCOLS
change. On the other hand emulator is more reliable than During the test it will be used a topology that consist of
simulator because emulator use real internetwork two end devices (PC_1, PC_2), two devices working on
operating system and emulate behavior of real devices. the third layer (GW_1, GW_2) and three routers (R_1,
Problem is that if we have complex topology we must R_2, R_3). All networking devices were manufactured by
have powerful device because emulation need lots of the company Cisco. The interconnection of individual
system resources. devices is accomplished by ethernet technology. Straight-
A. Optimized Network Engineering Tools through CAT5e cables were used as the transmission
medium. In Fig.3 is shown created topology.
Authors in [6] point out that OPNET is a commercial tool
designed to simulate the lifecycle of a computer network,
measure its performance, and identify its possible
vulnerabilities and network problems. Often, it is also
used for research purposes and the practical
implementation of developed protocols and technologies
before deploying them to the real environment. Accuracy
of results, rich functionality and simple user environment
Figure 3. Topology used for testing purpose
426
Authorized licensed use limited to: ULB Darmstadt. Downloaded on December 19,2022 at 21:32:52 UTC from IEEE Xplore. Restrictions apply.
ICETA 2018 • 16th IEEE International Conference on Emerging eLearning Technologies and Applications • November 15-16, 2018, Starý Smokovec, The High Tatras, Slovakia
In the topology above there are three possible paths 10,000 messages generated. The communication will be
between end devices PC_1 and PC_2. From the between the end devices PC_1 and PC_2.
perspective of dynamic routing protocols, each path is The same logical topology will be implemented in real
configured with different metric. Path through router R_1 environment, simulator OPNET and also emulator GNS3.
is the best and path through R_3 is the worst. Settings of hPing3 and used devices will vary because of
The Tab.1 contains configuration details of interfaces restriction in used software.
that influence metric of each path.
A. Topology in OPNET
TABLE 1. INTERFACES PARAMETERS In order to implement topology into simulator OPNET
we have to use objects that are presented in the Tab. 3.
OSPF (cost) EIGRP (delay)
Link TABLE 3. OBJECT IN OPNET
427
Authorized licensed use limited to: ULB Darmstadt. Downloaded on December 19,2022 at 21:32:52 UTC from IEEE Xplore. Restrictions apply.
ICETA 2018 • 16th IEEE International Conference on Emerging eLearning Technologies and Applications • November 15-16, 2018, Starý Smokovec, The High Tatras, Slovakia
EIGRP 0,000485
Figure 5. Topology created in emulator GNS3 EIGRP without FS 0,000835
OSPF 10
Object PC_1 is cloud that is used to connect topology and
devices in GNS3 with network card. This object was used The Tab. 4 contains the results obtained during
because in emulator there is not any device that can simulation of the convergence of dynamic routing
generate the amount of data that was designed for testing. protocols in the OPNET simulator.
For this purpose was used external device which have
installed hPing3 tool. B. Emulation in GNS3
V. EVALUATION In Fig. 7 are shown the data obtained by network analyzer
Wireshark during testing the protocol OSPF.
Testing the speed of convergence of dynamic routing
protocols has demonstrated the different behavior of the
protocols on different platforms but in every situation the
protocol EIGRP converges faster. The following sections
contain test results obtained on individual platforms.
A. Simulation in OPNET
Simulator OPNET provide the opportunity to have
complete control on topology during simulation.
Simulator does not provide only information about
convergence time but also about amount of traffic
generated by dynamic routing protocols which is depicted
in Fig. 6.
428
Authorized licensed use limited to: ULB Darmstadt. Downloaded on December 19,2022 at 21:32:52 UTC from IEEE Xplore. Restrictions apply.
ICETA 2018 • 16th IEEE International Conference on Emerging eLearning Technologies and Applications • November 15-16, 2018, Starý Smokovec, The High Tatras, Slovakia
429
Authorized licensed use limited to: ULB Darmstadt. Downloaded on December 19,2022 at 21:32:52 UTC from IEEE Xplore. Restrictions apply.
ICETA 2018 • 16th IEEE International Conference on Emerging eLearning Technologies and Applications • November 15-16, 2018, Starý Smokovec, The High Tatras, Slovakia
laboratory devices. Using the OPNET simulator and the [2] O. Bonaventure and P. Francois, Avoiding transient loops during
convergence of link-state routing protocols, ACM Transaction
GNS3 emulator showed significant flaws. Networking. IEEE, 2007, pp.1280–1292.
The GNS emulator provided more accurate results than [3] P. Palúch and N. Kocharians. CCIE Routing and switching. Cisco
the OPNET but a few shortcomings have been also Press, 2015. ISBN 978-1-58714-396-0.
identified. Based on the difference results of the tests [4] G. K. Dey, M. M. Ahmed, K. T. Ahmmed. Performance analysis
obtained in emulations and on real devices, it can be said and redistribution among RIPv2, EIGRP and OSPF routing
that the EIGRP protocol is more dependent on hardware protocol. 1st International Conference on Computer Information
Engineering. IEEE, 2015, pp.21-24, ISBN 978-1-4673-8342-4.
than the OSPF protocol. [5] T. Moy. OSPF: Anatomy of an Internet Routing Protocol.
Addison-Wesley Professional, 1998. ISBN 0-201-63472-4.
ACKNOWLEDGMENT
[6] A. S. Sethi, V. Y. Hnatysh. The Practical OPNET User Guide for
This paper is supported by grant project 011STU-4/2017, Computer Network Simulation. CRC Pre ss, 2012. ISBN
granted by the Cultural and Education Grant Agency of 978-1-4398-1206-8.
[7] H. Y. Z. Lu. Unlocking the Power if Opnet Modeler. Cambridge
the Slovak Ministry of Education. University Press, 2012. ISBN 978-0521198745.
REFERENCES [8] J.C. Neuman. The Book of GNS3: Build Virtual Network Labs
Using Cisco, Juniper, and More. No Starch Press, 2015. ISBN
[1] J. Carroll and J. Doyle. 2005. CCIE Professional Development 978-1593275549.
Routing TCP/IP. Cisco Press, 2005. ISBN 1-58705-202-4.
430
Authorized licensed use limited to: ULB Darmstadt. Downloaded on December 19,2022 at 21:32:52 UTC from IEEE Xplore. Restrictions apply.