Layerd software models for identification and refinement of performance requirements

Abstract

The primary measure of software system success is the degree to which it meets the purpose for which it is intended. In this view, requirement engineering is the process of discovering that purpose by identifying stakeholders and their needs. The requirements for a software system are often classified as functional and nonfunctional requirements. Functional requirements reflect the functionalities of the software system whereas nonfunctional requirements are the constraints on the services or functions offered by the system. Main nonfunctional requirements are performance, reliability, security and usability. Among these nonfunctional requirements performance is most important nonfunctional requirement that contribute maximum to the success or failure of the software product. This has four parameters namely response time, throughput, execution demand and resource utilization. Google search engine is the one real world working example which reflects the importance of performance requirements. It would not have become most successful if its response time is more and throughput is less. The second example is related to NASA. NASA had to delay the launch of a satellite for more than eight months because the flight operations segment software had unacceptable response times for developing satellite schedules. In the above applications response time and throughput played major role in success or failure, but in engineering and scientific applications CPU time plays major role whereas in embedded systems memory is very important. Due to gaining importance of performance requirements, in recent times, researchers are showing lot of interest in this area. Non-compliance of performance requirements causes to performance failures. This performance failure leads to performance risk. Performance failures cannot occur if all the performance requirements newlineix newlineare specified correctly and consistently. In this direction, layered models are proposed for performance requirements identification and refinement. These models can be used