Publication Date


Degree Program

Department of Mathematics and Computer Science

Degree Type

Master of Science


Software Engineering (SE) has been described as the discipline devoted to the design, development, and use of computer software, covering not only the technical aspects of building software systems, but also management issues develops highly complex software. The crisis in SE, due to the lack of well-defined formal processes, has led to poorly designed products with high maintenance costs and whose behavior becomes unpredictable. Component Based Software Engineering (CBSE) is currently a preferred approach to system design to overcome the crisis of SE, since it promotes software re-use, facilitates adaptability and faster system development. A component provides a function or a set of related functions, which forms a reusable program building block that can be combined with other components to form an application. A component with qualities such as, reusability, testability, modularity, complexity, proper to communicate and stability reduces maintenance costs. The components thus integrated, should be able to interoperate so that an operational application that results in reduced maintenance costs can be composed with minimal effort. Metrics are used to measure a component's quality factor and there are no good metrics available to validate their effectiveness, when components are integrated. Currently, the success of projects based on the CBSE methodology relies on experts who assess software components; however, their evaluation process involves parameters that may not be measured in practice. Existing traditional metrics are inappropriate since CBSE is aimed at improving interoperability and re-usability. Size metrics based on lines of code are not applicable as component sizes may not be known a priori. Furthermore, complexities that arise due to varying nature of facets and interfaces are not addressed by traditional metrics. This thesis addresses the evaluation of a series of metrics based on complexity, criticality and dynamic behavior, in order that component integration performance can be assessed. Three suites of metrics defined by various authors have been considered for evaluation so that one could choose the best metrics to measure an integrated environment. A suite of metrics proposed by Narasimhan and Hendradjaya are classified based on the attributes of: complexity, criticality and dynamic aspects. These metrics use graph-based connectivity to represent a system of integrated components. While the complexity metrics consider the packing density of integrated components and the interaction density among the components, criticality metrics reveal the extent of binding within each component in the system. Dynamic metrics have also been collected during the execution of an application and aid the process involved in testing and maintenance. Metric related data sets have been from several benchmark programs using instrumentation programs and key inferences have been obtained; these inferences include a systematic evaluation of quality of the various metrics. Two new metrics have also been provided towards assessing the stability of the application: one metric, namely CRIT instability, calculates the instability of each component, while the second new metric, namely CRIT inheritance,counts the number of components whose children exceeds a threshold value. Both these metrics are useful to assess the stability of the application and, in addition, to determine the components in a given application that needs to be redesigned. Future work will focus on the development of a metric evaluation suite to assess the system's stability as a whole, considering the role of each component in an application.


Computer Sciences