Student Projects in Computer Networking: Simulation

versus Coding

Leann M. CHRISTIANSON
California State University, East Bay
Department of Mathematics and Computer Science
Hayward, CA 94542-3092, USA

and

Kevin A. BROWN
California State University, East Bay
Department of Mathematics and Computer Science
Hayward, CA 94542-3092, USA

ABSTRACT

In this paper, we discuss the advantages and paper, we consider the pros and cons of using network
disadvantages of using network simulators to teach simulators to teach computer networking concepts versus
networking concepts versus having students write having students write programs. The authors have experience
programs. The authors have experience developing developing laboratory exercises for classroom use in both
laboratory exercises for classroom use in both realms [1-3]. Issues that impact the development of effective
realms. Two case studies, TCP versus UDP protocol networking assignments are presented along with case studies
performace, and routing algorithm convergence are and student feedback.
described. For both studies, students either wrote a
program that implemented the protocols being tested Computer networking is an integral part of computer
or created a simulation using the OPNET Modeler science education today. Creating practical and compelling
package. Issues specific to these projects are student assignments in networking, however, is challenging.
discussed. Student opinion reflected that simulation This is due to the fact that networking concepts and algorithms
software allows more focus on the performance of often involve many entities or are designed to work in
the protocol and visualization of results. environments that are not readily available for classroom use.
Programming, though viewed by students as more Some of the issues to consider when designing computer-
difficult, allows them to practice and gain skills that networking exercises are listed below:
will be useful in their future careers. The authors feel
that a balance of both programming and simulation a. Many protocols differ in how they respond to
activities is the best practice. corruption, errors, and loss, which occur
uncommonly and inconsistently in real networks
Keywords: Simulation, network, teaching, OPNET making replication of results difficult.
b. The performance of data-link layer algorithms may
only be apparent under high load which may
1. Introduction negatively impact other users of the network.
c. The differences in the performance of transport
When teaching computer networking it is algorithms may only be apparent if the transmitter and
important to design activities for students that receiver nodes are separated by large distances or if a
illustrate the concepts covered in lecture. In this large number of machines are linked to the network.
d. Routing algorithms require communication servers with which to interact (e.g., to test a student
with tens to hundreds of other nodes. web client).
e. Security concerns within a university may f. Super-user access may be needed to use commands
make it difficult to provide access to live that report on network performance
 g. Much network code is at the kernel-level generators may be used to emulate load, but inducing errors,
requiring recompilation of the kernel in loss and excess delay is a nontrivial enterprise. Since many
order to modify. algorithms require communication with many nodes over a
distance, they would necessarily rely on the traffic, error, and
Computer networking algorithms are complex, loss patterns seen in the Internet. These patterns are by no
therefore, it is beneficial to provide more than a means reproducible.
description and explanation of them in a lecture
setting. Students often learn more by experimenting Simulation software as an alternative to programming
on their own, by poking and prodding an algorithm to offers many benefits. Sophisticated simulation software
see how it reacts to different stimuli. The ability to allows for rapid construction of both wide and local area
experiment may be provided in (at least) two ways. networks, intermediary, and host nodes. Protocols can be
Students can be supplied with a simulation easily interchanged and tested. Large distances and multiple
environment, or they can be asked to implement an nodes can be simulated and errors, losses, and delay rates can
algorithm themselves and run it on real machines. be defined to follow a chosen distribution pattern. This allows
for reproducible results. Simulation software applications do
2. Simulation Versus Programming not require super user access or a dedicated lab. Students can
experiment without administrative fear of a compromised
In the computer networking area, as in much network. Timeliness is also a factor as long time periods
of Computer Science, it is not considered sufficient to (hours and weeks) can be simulated much more quickly than
simply teach the theory of a concept. Computer experimenting in real time.
Science may be considered engineering, and as such,
Computer Scientists should be able to apply what There are negative aspects of using simulation
they learn by building an artifact or program which software as well. Valuable class time will need to be spent
embodies the concept. In much the same way that a learning how to use the software. Some simulation
person who builds an automobile will have a better applications do not allow the user to modify all simulation
understanding of its workings than a person who only entities. This means that it might be impossible to demonstrate
knows how to drive it, our students will understand a particular event such as a pattern of movement of mobile
algorithms more deeply by implementing them. In nodes for example. Additionally, simulation results will
addition, undergraduate students should be given depend on the simulation model which may not accurately
every opportunity to improve their coding and widen reflect real traffic patterns. Cost, support, and license
the application of their skills. management of simulation software will also need to be
considered. Lastly, and most important, the students will not
There are many practical considerations in gain experience writing actual networking code and interacting
computer networking, however, which may preclude with live standards-based servers.
a project’s successful creation and testing in a real
environment. Even with a completely dedicated
network laboratory in which students are given super- 3. Case Study: TCP Versus UDP
user access, many experiments may not be conducted
due to the need for additional nodes at a distance to The effect of packet size on Transmission Control
act as correspondents. Without a dedicated lab, Protocol (TCP) and User Datagram Protocol (UDP) round-trip
students may be unable to modify network algorithms delay is a concept presented in most beginning network
residing in the kernel. Due to security concerns, courses. We have assigned this exercise as both a simulation
students may not be able to run network interfaces in and a programming assignment. Using simulation software
promiscuous mode in order to monitor packet arrivals such as OPNET Modeler [4], students can easily “drag and
on the local LAN. drop” networking components to a workspace to create a
network model. Components can be linked together using a
Testing algorithms presents another issue. variety of link speeds, and nodes can be configured to use the
Even within a dedicated lab, it is nearly impossible to TCP or UDP protocol via drop down menus. Figure 1
exactly duplicate the conditions that would be illustrates a wide area network with a File Transfer Protocol
required in order to evaluate an algorithm such that (FTP) client and FTP server. A base packet size and traffic
all students would see identical results. Traffic pattern can be configured from a list of options. The end
nodes can be configured with a drop-down menu to use TCP or
UDP packets when simulating a file request.

The simulation can be rerun with different packet

sizes, and results can be graphed for easy comparison.
Students should see that TCP packet delay is longer than UDP
packet delay due to the added complexity of the protocol.
Additionally, students should note that the delay to process
each TCP packet increases with packet size. Figure 2
illustrates a graph of TCP and UDP delay times for a serve to encourage the students to play with their simulation,
file transfer. discovering and learning on their own. A disadvantage,
however, is that the students will not gain valuable experience
in writing programs that use communication protocols. This is
certainly an important skill in which computer science students
should be accomplished

As an alternative, an instructor may instead choose to

have students write their own programs for this experiment [3].
This involves writing a client program and a server program
that communicate via TCP or UDP sockets, running multiple
tests, exporting results to a file and graphing them via Excel.
A description of the programming assignment is below:
TCP vs UDP: UDP is a simple best-effort, demultiplexor
protocol. TCP, on the other hand, is reliable and connection-
oriented. TCP, therefore, will go through a connection set up
phase prior to sending a packet. In this assignment, you will
compare delay values for each protocol. This can be
accomplished by writing 2 pairs of programs which will
exchange information using TCP and UDP and sockets.
Figure 1. OPNET model of a wide area The first pair of programs (a sender and a receiver) should use
network with an FTP client and FTP server. UDP to send a packet of set size back and forth between the
source and destination 10,000 times. You will want to
calculate how long the 10,000 round trips take, therefore, the
source must time the process. The source should check the
time when it first sends the packet and again check the time
when the packet comes back the 10,000th time. Next, your
program should calculate the difference, and use this
difference to calculate the delay for a single, one-way trip.
Note that the delay consists of processing delay, queuing
delay, transmission delay, and propagation delay. The last
three components of this delay should be the same for both
UDP and TCP (on the average). The second pair of programs
should do the same thing as the first pair, but use the TCP
protocol instead. Run each pair of programs 100 times and
create a histogram that displays the outcomes.

This assignment appears fairly straightforward, but

there are some issues that can make the testing complex.
Packet size can be adjusted easily, but it is difficult to generate
a consistent level of background traffic. Measuring results in a
real system is difficult as there is no way to control the actions
of other users competing for network resources. Programming
languages (C++ or Java) and operating systems (Unix or
Windows) will affect delay times, and results among students
Figure 2. Graph of TCP versus UDP packet delay for will vary. It is our experience that students become confused
a file transfer. when their results are inconsistent with what they have learned
in class. For instance, due to unseen circumstances, TCP delay
A benefit of this type of simulation is that
may appear to be less than that of UDP, contradicting what the
once the initial simulation has been completed,
experiment was meant to illustrate. Students will have trouble
students may change parameters such as link speed,
discerning whether the problem lies in their code or in the
packet size, and background traffic load to observe
network conditions at the time of the test. The programs will
how these affect their initial results. Various types of
also need to be run repeatedly, hundreds of times in order to
application background traffic such as web, database,
get a good test sample. This takes much longer (20 minutes
or email can be chosen. We have found that the
per run versus 1 minute) amount of time than an OPNET
ability to easily change factors which affect
simulation. Finally, a big disadvantage is the lack of
performance, the quickness in which each simulation
immediate graphical output. Program results will need to be
runs, and the ability to view graphically the results all
output to a file and later imported to a spreadsheet for
graphing. Many students have trouble with this and understanding of routing table convergence and its
forego it all together. implications.
Student comments reflect that the
experiment’s goals and outcomes are clearer when
simulation software is used. Students enjoy playing
with the simulation and tweaking model components.
The students do feel, however, that the experience of
coding and using sockets for program communication
is worthwhile. For many, this is the first time that
they have been exposed to socket programming.
Though students do enjoy flexibility, some problem
issues of this exercise can assuaged by having all the
students use the same programming language,
operating system and network.

4. Case Study: Routing

Convergence
Routing algorithm convergence time is
another lesson that students can explore through
simulation or programming. Convergence time is the
time it takes for routes within a routing table to
stabilize after start up or after a link has gone down.
Common routing algorithms such as Open Shortest
Path First (OSPF), the Routing Information Protocol
(RIP), or others can be used. Using OPNET Modeler
to simulate this exercise is straightforward. A group Figure 3. Simple topology for routing convergence.
of routing nodes can be connected into a network of a
given topology. Figure 3 illustrates a simple
topology with five nodes. Each node can be
configured to run the routing protocol of choice.

OPNET provides the capability to export

routing tables to a file as well as to allow the student
to “break a link” after a certain amount of simulation
time. Students can then study the routing tables to
find the new path that was propagated to the routing
tables after the link went down. This exercise gives
the student experience with router interfaces and
interpretation of routing tables. Students can then
study the simulation and routes and calculate the
convergence time for the original network and then
again for the network after one link goes down.
Figure 4 illustrates a routing table for node 0 at 8
seconds into the simulation.

The advantage of using a simulation for this study is

again, rapid configuration and design. The routing
protocol, number of nodes, and interfaces can all be
configured easily. Results can be viewed graphically Figure 4. Routing table for node 0 after 8 seconds of
in a quick and easy manner, and routing tables are simulation time.
automatically created. Students enjoy creating a
simulation for this exercise. Many have stated that Coding a program that simulates a routing protocol,
they understand the routing concepts discussed in creates routing tables, and allows a user to follow a route from
lecture much better after having completed this one node to another is a rather complex task. An input file of
assignment. In particular, they feel that they have a nodes and weights that will represent the network topology
better understanding of routing table organization, will need to be provided by the instructor. The student’s
and how to interpret the routes created by a particular program should read this file, simulate the routing algorithm,
routing protocol. They have also gained increased
and create routing tables. The students will need to computer networking courses. Two case study
write another program to follow the routing tables
from a source node to a destination node to confirm
the performance of their program. A description of
the programming assignment follows:
The Distance Vector Routing algorithm was
one of the first routing algorithms used on the
Internet and is part of the RIP protocol. For this
assignment, you will need to write a program which
simulates the behavior of the Distance Vector
Routing Algorithm. Details are below: Assume that
there are 5 nodes in the graph. You will be supplied
with an input file which will have a triple <nodeA>
<nodeB> <weight> on each line. This will indicate
the links between nodes of the graph and their
weight. Assume that links are bi-directional with the
same weight in each direction. If a link does not
exist, assign it a weight of infinity. To begin you will
need to read this file and store the information
regarding the initial shape and link values of the
graph. Next, you will need to run the Distance Vector
Routing algorithm until it converges (routes stop
changing).

After convergence, your program should

print the following: the routing table for each
node,the number of cycles it took to converge. Your
program should then query the user for a source and
destination node. Given these 2 nodes, you should
then inspect your routing tables and print out the
route from the source to the destination and the
weight of that route.

From our experience, the disadvantage of

coding this particular problem is that the problem is
complex enough that it will take a good investment of
student time in order to complete the project.
Students often get bogged down with the coding
details and lose sight of the purpose of the
assignment. Many will not be able to write programs
that generate correct results. One technique that we
find aids success is to convince the students to draw
the graph and run the routing algorithm by hand prior
to coding. This helps them visualize what their
program needs to do and thus makes design and
coding a bit easier. Students who complete this
project successfully have recounted that they feel
they have gained coding experience, but have not
gained any insight on routing and convergence. For
this particular project, a simulation might appear to
be better at teaching the concepts. The simulation
allows the students to focus on the main points of the
lesson, routing and convergence, without getting lost
in the coding details.

4. Conclusion
We have discussed the advantages and
disadvantages of simulation versus programming for
activities were described: the first, compares the
round-trip packet delay time for both TCP and UDP
protocols, the second looks at routing protocol
convergence time. For both studies, students either
wrote a program that implemented the protocols
being tested or created a simulation using OPNET
Modeler. Student opinion reflects that simulation
software allows more concentration on the
performance of the protocol and allows for an easier
visualization of results. With simulations, students
are more apt to “play” with their simulations and
experiment beyond the bounds of the assignment.
Programming, though viewed by students as more
difficult, allows them to practice and gain skills that
will be useful in their future careers. The authors feel
that a balance of both programming and simulation
activities is the best practice.

5. References
[1] Christianson and Brown, OPNET Lab
Manual to Accompany Business Data
Communications by William Stallings, 5th
Edition, Pearson Prentice Hall 2005.
[2] Brown and Christianson , OPNET Lab
Manual to Accompany Data and
Computer Communications 7th Edition
and Computer Networking with Internet
Protocols and Technology, 4th Edition, by
William Stallings, 5th Edition, Pearson
Prentice Hall 2005.
[3] Brown and Christianson, Networking
Lab Exercise for The Networking and Data
Communications Laboratory Manual,
edited by Frances S. Grodzinsky, published
by Prentice Hall.
http://www.prenhall.com/grodzinsky.
[4] OPNET IT Guru http://www.opnet.com