|
Abstract:
|
The use of software is gaining ground from traditional hardware-based systems development. At the same time, competition becomes more intensive and new products are introduced to the market at an ever-increasing pace. This creates the need for improved prediction and management of software projects. One of the approaches that has been proposed as an aid in software development is the use of quantitative techniques, often referred to as `metrics'. The purpose of this Master of Science thesis is to investigate the field of quantitative methods available today, especially from the viewpoint of project control. The aim is to identify useful techniques that can be applied in a commercial organization developing various types of software with real-time characteristics. The thesis illustrates the basic principles of quantitative techniques and discusses the meaning and classification of software measures. The role of measurement in software development and project management is also examined, and a framework for analysing metric data-is presented. The difficulties involved in measuring software are illustrated from several viewpoints. The most important metrics and quantitative methods are examined; these include project cost estimation as well as measurement of the properties of intermediate products. The thesis demonstrates various problems with these metrics and concludes that the development of software metrics has been partly based on unrealistic expectations. In many cases simplistic and easily extracted measures have been developed without due consideration of the complex human activities that they should model. Consequently, these metrics have little theoretical basis or empirical support. It nonetheless seems that software organizations should monitor their activities more closely, but using the measurement data for prediction places stringent requirements on the consistency of the development process. The thesis finishes by proposing a simple set of measures for monitoring project progress and the quality of work products. /Kir10 |