Engineering: Working Papers
http://hdl.handle.net/10092/612
2016-08-30T07:19:23ZQueueing Models With Finite and Infinite Buffering Capacity : A Comparative Study
http://hdl.handle.net/10092/12544
Queueing Models With Finite and Infinite Buffering Capacity : A Comparative Study
Anggawijaya, Hermin
Queueing theory offers a large variety of techniques that can be used in performance modelling of computer systems and data communication networks. The diversity of assumptions causes that the numerical results, obtained for the same system but by means of different techniques, can often be numerically very different. The project is aimed at finding common denominators for numerical results obtained with the assumption of infinite buffer capacity and that obtained with the assumption of finite buffer capacity. To be precise, this project investigates the traffic intensity regions where the approximation of queueing systems with finite buffer capacity by the queueing systems with infinite buffer capacity can be done with some amount of safety margin. The investigation includes both queueing systems with single arrivals (M/M/1, M/D/1), and a queueing system with batched arrivals (M⁽ᵇ⁾ /M/1), and two the most important characteristics of queueing systems are considered, the probability of overflow and average waiting time in the system.
1991-01-01T00:00:00ZOptimising hashing functions with genetic algorithms
http://hdl.handle.net/10092/12543
Optimising hashing functions with genetic algorithms
Botting, Mark
Genetic algorithms (aka GA's) are a robust global search strategy that ignore local minima and irrelevant parameters, suitable for large search spaces. It is based on an analogy with natural evolution and survival of the fittest. A generation of potential solutions is formed by randomly mating pairs from the previous generation, giving preference to the better performers. By repeating the process many times a (near) optimal solution evolves.
Hashing functions are an implementation for fast table lookup, searching, etc. Given a symbol to store or lookup in a table, a hashing function produces an index into the table, preferably such that all possible symbols will be evenly distributed throughout the table. Their effectiveness is controlled by various parameters such as table size, symbol distribution, and the form of the function itself.
This project aims to couple these two areas together to optimise the parameters of a given hashing function, by searching for a good set of values for the parameters with a genetic algorithm.
1991-01-01T00:00:00ZMawd: A Collaborative Analysis Tool
http://hdl.handle.net/10092/12542
Mawd: A Collaborative Analysis Tool
Macfarlane, Justin Ray
Video analysis is a powerful tool for studying human interactions. Transcribing video footage into usable and understandable data is tedious and time consuming. This report gives an overview of evaluation in RCI and CSCW, and describes the design and implementation of a Collaborative Video Analysis tool that eases data extraction and evaluation.
1995-01-01T00:00:00ZCeramic tiling viewer
http://hdl.handle.net/10092/12447
Ceramic tiling viewer
Burt, Landery
Observing and interacting with multi-thread programs can be difficult for the programmer. Simple input/output (I/O) can become a nightmare when multiple threads read and write simultaneously. A solution would separate the I/O streams of the multiple threads, windowing techniques can achieve this.
This honours project report presents the design and implementation of Ceramic, a development tool which assists in observing and interacting with multi-thread programs. Multiple viewers (windows) can be opened to control I/0 streams of multiple threads. Ceramic has an object-oriented design based on design patterns captured from Mossenbock's OberonO viewer system. Another feature are the hierarchical tiling viewers which are a hybrid of Elastic Windows developed by Kandogan & Shneiderman. Tiling viewers have some significant advantages over overlapping windows which Ceramic has exploited.
1996-01-01T00:00:00Z