Local area networks (Computer networks)

This thesis investigates the waiting-time in a token-ring local area
network (LAN) with gated service discipline. It gives an analytical
approximati.on for a network with an arbitrary number of stations,
non-identical Poisson arrival streams and general service time
distributions at each station, and non-zero switch-over time (also
referred to as walking time) between adjacent stations. This
approximation method is an alternative to the complicated O(N^2) numerical evaluation of the waiting-time currently available in the
literature. Extensive simulation results with different parameters are
presented to show the degree of accuracy of the approximation, which is
generally good for a practical range of parameters.

This thesis gives an approximate analytical solution in terms of mean
end-to-end delay of a system of token-ring local area networks
interconnected via a backbone ring. The solution is based on an existing
approximation (see chapter 2) of the mean end-to-end delay in a
stand-alone token ring LAN, and it is extended by approximating the
arrival rates at the bridges (that are function of the throughput of each
subnetwork and the percentage of it that is directed to the other
subnetworks). A new packet-length distribution is defined and its
detailed derivation is described in Appendix B.

This thesis deals with simulation of a conflict-free token-passing protocol
for local area bus networks. The primary emphasis of this simulation study
is to observe the effects of token holding time on the performance of the
network. Token holding time is adjusted to account for three types of
service disciplines: purely non-exhaustive, non-exhaustive and exhaustive.
Network performance for these three service disciplines is compared to
determine, which one of the three gives a relatively better performance.
Besides throughput and delay, a more compact form of performance measure
called "power", has also been used in this study. Power is simply a ratio
of throughput and delay. This study has shown that the token holding time
has significant effect on the performance of a local area network.
Simulation results of this study, are presented in terms of throughput,
delay, power, logical ring size, token circulation time,
efficiency/overhead versus offered load and token holding time. Some
results are also presented in terms of histograms.

In this dissertation, an object-based I/O architecture for personal computers (PCs) and workstations is proposed. The proposed architecture allows the flexibility of having I/O processing performed as much as possible by intelligent I/O adapters, or by the host processor, or by any processor in the system, depending on application requirements and underlying hardware capabilities. It keeps many good features of current I/O architectures, while providing more flexibility to take advantage of new hardware technologies, promote architectural openness, provide better performance and higher reliability. The proposed architecture introduces a new definition of I/O subsystems and makes use of concurrent object-oriented technology. It combines the notions of object and thread into something called an active object. All concurrency abstractions required by the proposed architecture are provided through external libraries on top of existing sequential object-oriented languages, without any changes to the syntax and semantics of these languages. We also evaluate the performance of optimal implementations of the proposed I/O architecture against other I/O architectures in three popular, PC-based, distributed environments: network file server, video server, and video conferencing. Using the RESearch Queueing Modeling Environment (RESQME), we have developed detailed simulation models for various implementations of the proposed I/O architecture and two other existing I/O architectures: a conventional, interrupt-based I/O architecture and a peer-to-peer I/O architecture. Our simulation results indicate that, on several different hardware platforms, the proposed I/O architecture outperforms both existing architectures in all three distributed environments considered.