ABSTRACT During recent years, the trends in computer resource organization have seen the replacement of the traditional mainframe computers and their timesharing terminals with individual desktop workstations. These workstations can offer more computational power than the mainframes of less than a decade ago. With this distribution of processing from the more traditional mainframe computer comes the need to share peripheral resources among these desktop workstations. These include items such as file systems, tape drives, printers, plotters, and graphics display devices. A common way to share these among the multi-computer environment is to provide a transparent gateway that allows each workstation to function as if it had a dedicated set of peripherals. As educational institutions and industry continue the trend of connecting their computer systems in networks, a serious hole has developed - the ability to estimate the throughput of their network. This paper describes the development, use, and verification of the Local Area Network Simulation Package (LANSIM). LANSIM provides a means to determine the performance characteristics of any network adhering to the ANSI/IEEE Std 802.3-1985 for local area networks. This will allow anyone to simulate a proposed network and gain insight into such crucial topics as percent utilization, aggregate data rates on the network, and response times - all prior to the purchase and installation of the network.

In this thesis, the design and performance issues of a protocol for a dynamic hierarchical local area network are addressed. The dynamic hierarchy is an extension of the static hierarchical topology with the exception that the apex node can vary among a prescribed set of candidates. A simulation model is used to evaluate the benefits of dynamic apex transition under the proposed protocol. The design also considers routing and queue structures of the nodal functions corresponding to the transport/network layers of the ISO model. The model experimentation shows that dynamic hierarchy networks are advantageous, under an environment where different network traffic patterns exist at different points of time, over static hierarchy networks. Critical to the conclusion is that the amount of time taken to effect a transition is kept minimal. / M.S.

Network management and control of large broadband local area networks, where the number of nodes ranges from several hundred to a few thousand, is a very important concern for today's network manager. This primarily involves tuning the network to ensure load balance over the broadband subchannels, and the capability to monitor specific nodes for accounting, performance and security purposes. This thesis presents the design considerations for a session level passive hardware monitor on a broadband local area network having a bus/tree topology. A methodology for session monitoring is presented based on the issues discussed. The session service provided is packet switched with point to point virtual circuit connection. Current technology and economics dictate the use of broadband transmission media for large local area networks spread out over a radius of around ten kilometers. This medium provides adequate throughput for a large number of devices by supporting frequency division multiplexing and a multiple access medium access control protocol. The design considerations include both hardware and software aspects and are justified based on the characteristics of the transmission medium and communication protocol architecture used in this study. The local area network used for the development of this project is a sixteen hundred node campus network at Virginia Tech (™LocalNet 20) supplied by SYTEK, Inc. / M.S.

The research described in this thesis deals with effective protocol specification. The primary question addressed is whether the Express Transfer Protocol (XTP), a "real-time" Transport layer protocol, is sufficiently specified or are there "holes" in its specification? A new protocol evaluation process is formulated and applied to XTP in order to answer this question. The evaluation process combines a detailed analysis of the XTP specification with an attempt to implement parts of the protocol from the specification. Special attention is given to those aspects of the protocol that affect "real-time" naval tactical communications. The detailed analysis of the specification and its effect on the specification revision process are presented. The analysis is described in formal comment papers and electronic mail transmitted to the protocol designer, Dr. Greg Chesson. Elements of the protocol most applicable to tactical communication are selected for implementation. A number of assumptions are made so that an implementation can be built. The design of the prototype implementation and the assumptions made to build it are discussed. Both the hardware and software being used to build the implementation are presented. The protocol evaluation process is found to be appropriate for evaluating XTP. A comparison is made between this technique of protocol evaluation and existing techniques (i.e., simulation, complete implementation, and protocol verification). The principal conclusion of the protocol valuation process is that a number of areas of the XTP Protocol Definition (prior to Revision 3.3) are not sufficiently specified. These areas must be properly specified before a complete implementation can be built. Without a complete specification, the communications support intended for XTP is unlikely to be realized. / Master of Science

The inherent properties of optical fibers such as small size and weight, EMI/RFI immunity, low attenuation, and large bandwidth provide many advantages over wire conductors that make fiber well-suited for communications. Fiber optic local area networks are particularly suited for use in electrically noisy and space sensitive industrial environments. The diversity of communication requirements that exist in a typical factory situation can be accommodated by the use of a hierarchical communications structure consisting of multiple tiers of fiber optic networks. The lowest tier of this structure would be inexpensive feeder networks used to connect devices such as sensors, actuators, PLCs, robots, and small computers on a factory floor. The emphasis at this level is low cost, but while providing interconnection to higher tiers. An approach which satisfies the lowest tier requirements is a non-shared medium scheme which is link based, consisting of an active star architecture using a roll-call polling access method. The centralized intelligence structure of a master/slave access method allows one to concentrate on the cost optimization of the optical data links. The use of low-cost optical components such as LEDs, PIN diodes, and plastic fiber as well as the potential for a significant amount of common hub equipment provides considerable economies. / Master of Science

No description available.

One of the local area network medium access control standards created by the Institute of Electrical and Electronic Engineers in IEEE Project 802 is the carrier sense multiple access with collision detection (CSMA/CD) medium access control. Numerous studies have been reported in the literature on the performance of CSMA/CD. These studies show that CSMA/CD performs well under light network load but not well at heavy load. To improve the performance of CSMA/CD under heavy load, a new concept called network partitioning is presented. Network partitioning allows the network to be partitioned into segments when under heavy load. Partition stations then act as bridges between the segments. The impact of network partitioning on network performance was tested using a simulation model of an Ethernet local area network (an implementation of the CSMA/CD medium access control). The simulation results show that network partitioning can improve the performance of CSMA/CD under heavy load.

Network research experiments have traditionally been conducted in emulated or simulated environments. Emulators are frequently deployed on physical networks. Network simulators provide a self-contained and simple environment that can be hosted on one host. Simulators provide a synthetic environment that is only an approximation of the real world and therefore the results might not be a true re ection of reality. Recent progress in virtualisation technologies enable the deployment of multiple interconnected, virtual hosts on one machine. Virtual hosts run real network protocol stacks and therefore provide an emulated environment on a single host. The rst objective of this dissertation is to build a network emulator (viNEX) using a virtualisation platform (XEN). The second objective is to evaluate whether viNEX can be used to conduct some network research experiments. Thirdly, some limitations of this approach are identified / Computing / M. Sc. (Computer Science)

No description available.

This thesis is concerned with the design of a high-speed backbone network which provides a high bandwidth interconnection for various Personal Computer Networks (PC-Networks) with an integrated service of voice and data. With the advanced technology of optical fiber as the transmission medium, several different existing topologies and protocols are discussed for the backbone network design. The token ring protocol is simulated and evaluated to find out a suitable buffer size and the length of voice and data packet for backbone network. The Network II.5 simulation tool is applied to simulate the token ring simulation model with different parameters. The Network Interface Unit (NIU) is designed from the simulation results with a cost-effective consideration.