360 likes | 741 Views
Student Projects in Computer Networking: Simulation versus Coding. Leann M. Christianson Kevin A. Brown. Cal State East Bay. Cal State East Bay. Introduction. Computer networking is an integral part of computer science education Network protocols involve many entities
E N D
Student Projects in Computer Networking: Simulation versus Coding Leann M. Christianson Kevin A. Brown Cal State East Bay
Cal State East Bay Introduction • Computer networking is an integral part of computer science education • Network protocols involve many entities • Creating practical and compelling student assignments in networking is challenging
Cal State East Bay Challenges… • Protocols differ in how they respond to corruption, errors, and loss, and replication of results difficult • Algorithm performance may only be apparent under high load, negatively impacting other users of the network
Cal State East Bay Challenges (cont’d)… • The differences in the performance of transport algorithms may only be apparent if the transmitter and receiver nodes are separated by large distances or if a large number of machines are linked to the network • Routing algorithms require communication with tens to hundreds of other nodes
Cal State East Bay Challenges (cont’d)… • Security concerns within a university may make it difficult to provide access to live servers with which to interact • Super-user access may be needed to use commands that report on network performance • Much network code is at the kernel-level requiring recompilation of the kernel in order to modify
Cal State East Bay NetworkingAlgorithms are complex • It is beneficial to provide more than a description and explanation • Students learn by experimentation • The ability to experiment may be provided in (at least) two ways
Cal State East Bay ExperimentationTwo Methods • Provide a simulation environment • Allow the students to implement the algorithm themselves and run it on real machines
Cal State East Bay Coding Benefits • Computer Science is similar to engineering • Computer Scientists need experience applying what they learn by building workable modules or programs that implement a key concept • Coding skills enhance career opportunities
Cal State East Bay Coding Issues • Dedicated equipment is needed • For security • Multiple nodes at great distances must be available • Testing • Difficult to duplicate conditions such that all students see identical results • Inducing errors, loss and excess delay is difficult even with a traffic generator • Internet traffic, error, and loss patterns are not reproducible
Cal State East Bay Simulation Software Benefits • Allows for rapid configuration - WANs, LANs, intermediary, and host nodes • Protocols can be easily interchanged and tested • Large distances and multiple nodes can be simulated • Errors, losses, and delay rates can be defined to follow a chosen distribution pattern • Results are reproducible • Dedicated equipment and super user access not required • Long time periods can be simulated quickly
Cal State East Bay Simulation Software Issues • Requires investment of class time • Simulation model may not accurately reflect real traffic patterns • May be difficult to reproduce a particular event as some simulation applications do not allow the user to modify all simulation entities • Students will not gain experience writing actual networking code and interacting with live standards-based servers • Cost, support, and license management must be considered
Cal State East Bay Simulation SoftwareOPNET • University Program provides • Free IT Guru software for each student • Runs on Windows XP • Unix version for university • www.opnet.com
Cal State East Bay Example 1TCP vs UDP packet delay • Simulate client/server communication over an IP network • Observe the effect of packet size on delay for both TCP and UDP transport layer protocols for file transfer
Cal State East Bay OPNET: FTP client and server
Cal State East Bay TCP vs UDP Results • Simulation can be rerun with different packet sizes • Results can be graphed for easy comparison • TCP packet delay is longer than UDP packet delay due to the added complexity of the protocol • Delay to process each TCP packet increases with packet size
Cal State East Bay TCP vs UDPFTP: Packet Delay
Cal State East Bay TCP vs UDP Simulation Benefits • Students can easily change parameters such as link speed, packet size, and background traffic load and observe how these affect their initial results • A variety of background application traffic (web, database, or email) can be chosen • Easy to view and compare results graphically
Cal State East Bay TCP vs UDP Simulation Disadvantages • Students do not gain valuable coding and testing experience
Cal State East Bay TCP vs UDP Coding • Write a client and server program that sends a small packet back and forth via a TCP socket 10,000 times • Run program 100 times, noting the round trip time for a round of 10,000 sends • Graph delay times • Repeat with a pair of programs using UDP sockets
Cal State East Bay TCP vs UDP Coding Benefits • Provides experience with socket programming • Often a first for CS majors • Allows students experience with the software lifecycle • develop, run, test, and document results
Cal State East Bay TCP vs UDP Coding Disadvantages • Difficult to generate a consistent level of background traffic • Inability 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 • Takes additional time to import results to Excel and create graphs
Cal State East Bay Example 2Routing Convergence Time • Convergence is the time it takes for routes between network nodes 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
Cal State East Bay RIP Routing ConvergenceSimulation • Create a network of router nodes configured to use a particular routing algorithm (RIP) • Evaluate the routing tables created by the routing algorithm and trace a path between two nodes. • Break a link and run the simulation again • Determine the routing table convergence time after a link is broken
Cal State East Bay OPNET: Routing Convergence
Cal State East Bay Routing Table: node 0 after 8 sec
Cal State East Bay RoutingSimulation Benefits • Rapid and easy configuration • Routing tables are automatically created • Can “break” links and rerun the simulation easily • Simulation runs quickly
Cal State East Bay Routing ConvergenceSimulation Disadvantages • Students do not gain experience implementing a routing algorithm on their own • Difficult to calculate the exact convergence time due to timing of simulation events • Some routing protocols are not offered
Cal State East Bay Routing ConvergenceCoding • Provide an input file with nodes, distances, and weights for initial topology • Implement a routing algorithm on the topology and run the program until convergence • Output routing tables after convergence • Write an additional program to trace routes through the tables
Cal State East Bay Routing ConvergenceBenefits • Substantial assignment provides a great coding experience • Allows students experience with the software lifecycle • develop, run, test, and document results • Students who complete the project successfully understand convergence and the way the routing protocol works
Cal State East Bay Routing ConvergenceDisadvantages • Problem is complex and requires a large time investment • Students get bogged down with the coding details and lose sight of the purpose of the assignment • Many give up or fail to write programs that generate correct results
Cal State East Bay Conclusions • Simulation software allows more concentration on the performance of the protocol and allows for an easier visualization of results • Students are more apt to “play” with their simulations and experiment beyond the bounds of the assignment
Cal State East Bay Conclusions • Programming, though viewed by the 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
Cal State East Bay THANK YOU! • OPNET Lab Manuals • Brown & Christianson, Lab Manual for Data and Computer Communications 7/E by William Stallings,Prentice Hall, 2005 • Christianson & Brown, Lab Manual for Business Data Communications 5/E by William Stallings, Prentice Hall, 2005 • Brown & Christianson, The Networking and Data Communications Laboratory Manual, Labs 5-8, edited by Frances Grodzinsky, Prentice Hall, http://www.prenhall.com/grodzinsky.
Cal State East Bay Contact Us… • Leann.Christianson@csueastbay.edu • Kevin.Brown@csueastbay.edu