Teaching Computer Networks Through Modeling Teaching Computer Networks Through ModelingbyCapt. Noël Davis, Capt. Scot Ransbottom and Lt.Col. Drew HamiltonElectrical Engineering and Computer ScienceUnited States Military AcademyWest Point, New York 10996INTRODUCTIONModern computer scientists need to understand both the theoretical and practical aspects of computer networking. As computer science teaching methods continue to mature, experiences in mixing theory and application have been shared in the community [Marti 96]. While the literature is rich with experiences in applied or project based courses in artificial intelligence, architecture, software engineering, data structures and algorithms, few have addressed the challenge of coordinating theory and practice in computer networking courses. Combining theory and practice in a single course on computer networks is difficult because of the complexity and scale of modern networks. Classroom modeling of networks is a technique to illustrate the theoretical aspects of networking through practical models of computer networks. At West Point we are evolving our network course to use network modeling and simple simulation as a visualization tool to show how networks operate and the performance tradeoffs associated with network design decisions. This approach provides the opportunity to do more than teach an undergraduate survey course in computer networking. RATIONALE FOR ACADEMIC NETWORK MODELINGMost networks are built to accommodate the needs of a single organization or group. Internetworking is a technology that accommodates multiple, diverse, underlying hardware by providing the means of interconnecting heterogeneous networks. [Comer 95]. Comparatively few individuals, academic institutions or corporations have networksexclusively dedicated for student use and experimentation. Prudent administrators limit student or general user access to operational networks. Modeling efforts are not limited by these operational considerations.Network courses are often based one or more of the following areas: - The OSI Model- Performance Analysis- Network SimulationAs we summarize each area, we must point out that most networks courses will cover most of these areas and but may widely vary the degree of coverage. OSI Model - based instruction is very common. Typically, the functionality of each layerof the OSI Model is explained and then examples of this functionality in a specific ACM SIGAda Ada Letters, Volume XVIII , Issue 5 Sept./Oct. 1998, pp 104 - 110Teaching Computer Networks Through Modeling network protocol is presented. However, there are few, if any, implemented network protocols in which the architecture layers are strictly aligned with the OSI model layers. Another way to approach the model technique is to teach the basics of a protocol then describe how it reacts at each level of the OSI model, if it engages a particular level at all.Performance Analysis - an emphasis on analytical models of networking and the associated probability is also challenging and worthwhile. Not all undergraduates will have the necessary math/statistical background for serious work in analytical modeling ofcomputer networks. While the academic value of analytical modeling is undoubted, it should be noted that an increasing number of researchers are questioning how well classical queuing theory approximates actual network performance [Paxson 95]. Simulation of computer networks provides students with simulated observations of the operation of a network. [Barnett 93] proposes the use of the NetSim simulator to support both major project assignments and more focused homework assignments. We have also found NetSim valuable in allowing students to observe and experiment the effects of bottlenecks, delays and other network phenomena. NetSim is available via anonymous FTP from lcs.mil.edu and runs over X-Windows. Whatever the simulation tool, a prudenttechnique is to incorporate the tool into supervised lab/project assignments, individual homework assignments, and classroom demonstrations.USE OF OPNETThose institutions that do have computer network test beds often find the maintenance and operation of such facilities to be manpower intensive. For example, once a test network is set up and operational, how is it stressed? How are comparisons between design decisions made? How many labs can purchase the newest ATM switches with the latest features?Network modeling does not suffer from these shortcomings. The U.S. Army has adopted OPNET as a standard under the auspices of the Army Enterprise Strategy under the leadership of the U.S. Army Office of the Director of Information Systems for Command,Control, Communications and Computers. OPNET is widely used in universities as well as many parts of the DOD. Our preliminary work with OPNET indicates that it will support our visualization through modeling objectives. OPNET may be described as a communications-oriented simulation language. The name OPNET is derived from Optimized Network Engineering Tools. The single most significant aspect of OPNET is that it provides direct access to the source code coupled with an easy-to-use front end. A generic approach to network modeling can be constructed using the OSI Reference Model as its basis. This approach allows the implementation of different network protocols which are compatible at the OSI layer boundaries. Pedagogically, this approachhas limitations. As illustrated in Figure 1 below, any detailed implementation of an Ethernet model will not directly align with the OSI Reference Model. Other protocols ACM SIGAda Ada Letters, Volume XVIII , Issue 5 Sept./Oct. 1998, pp 104 - 110Teaching Computer Networks Through Modeling such as Fiber Distributed Data Interface also do not perfectly align with the OSI Reference Model. Figure 1. Reference Models and ArchitecturesOPNET models are composed of three primary model layers: the process layer, the node layer and the network layer. The lowest modeling layer is the process layer.Network Model snetwo rksandsubnetwork sNodeModel sindividualnodes andstationsProcessModel sSTD thatdefin es anodeFigure 2 OPNET model hierarchy.This modeling hierarchy is illustrated in Figure 2. The process model in Figure 3 shows a state transition diagram (STD) for the generation of packets. Process models are built ACM SIGAda Ada Letters, Volume XVIII , Issue 5 Sept./Oct. 1998, pp 104 -