1 Metrics for Software

Objective measurement of non-functional requirements (NFRs) in any software is one of the most difficult activities. This is because the inherent nature of the NFRs makes their common understanding difficult. The problem is compounded by the fact that the requirements for any software are usually vague about the NFRs that the software should satisfy and how to evaluate the NFRs in the final software. An ideal solution to these problems will be the development and usage of metrics for all NFRs – this will let the software be checked for the NFRs at all stages of its lifecycle and in case of deviation from the requirements corrective action may be taken. While there are several NFRs such as performance, maintainability, reusability, security, and so on, among the more important of the NFRs is adaptability. Intuitive definition of adaptability is the extent to which a software system adapts to change in its environment. An adaptable software system can tolerate changes in its environment without external intervention. For example, a dual-mode cell phone can find out by itself if any one of the two wireless standards it supports is available at its current location and if so it starts using that standard. A practical metric for this NFR will help software developers and their organizations. This paper discusses our initial work in developing metrics for adaptability. In this paper we have developed several metrics for software adaptability. The advantage is that the metrics that we have developed are applicable at the architectural level. Since architecture development is the first stage of the design process, the extent to which the architecture is adaptable will determine the adaptability of the final software. Hence the metrics in this paper will help determine the extent to which the final software will be adaptable as well. This paper first defines software adaptability and then defines adaptability indices – the architecture adaptability index (AAI) and the software adaptability index (SAI). The application of these indices is illustrated with examples. Also the validity of these indices for hierarchical architectures, legacy systems and for dynamic adaptation is examined.