Eliciting Small Scale Architectures with Maintainability Scenarios

Abstract

There are available methods for software architecture elicitation and evaluation. Many of them have been applied in industry with great success. These methods are well suited to bigger operations, and there have been many successful applications in large scale software production. These successful methods are less effective for small to moderate scale software production in comparison. In addition, existing elicitation methods frequently consider several quality attributes. In some cases, especially in smaller production, a specific quality attribute is the primary concern. There are few elicitation methods available that are designed for that context. This study proposes an elicitation method designed for small to moderate scale software projects where achieving high maintainability is the primary concern. This method was named the Maintainable Architecture of Small Scale Elicitation Method or the MASSE method. This method consists of three phases each of which is comprised of several steps. These phases are executed iteratively until a satisfactory result is obtained. The phases are Presentation, Elicitation and Execution. During the Presentation phase the MASSE method is explained to relevant parties, during the Elicitation phase the requirements for the software architecture are elicited and during the execution phase the architecture is designed based on these requirements. The method utilizes Quality Attribute Scenarios as its main elicitation tool. To test the effectiveness of the MASSE method it was applied in industry. The method was applied at an Icelandic engineering firm called EFLA, with relatively small software operations. There an existing program called EflaVefgatt needed to be redesigned to substantially increase its maintainability. Two iterations of the method were executed during which both the architecture design and the method were improved. Qualitative data was collected for both iterations through interviews. The interviews compared the existing software architecture with the architectures resulting from the two iterations of the MASSE method. The results were positive after the first iteration and after the improvements made during the second iteration they were more positive and deemed satisfactory. Subsequently the process ended as a success.