•   When: Friday, April 29, 2016 from 09:00 AM to 11:00 AM
•   Speakers: Ehsan Kouroshfar
•   Location: Nguyen Engineering, Room 4201
•   Export to iCal

Abstract

Conventional wisdom suggests that a software system's architecture has a significant impact on its evolution. Well-designed software architecture employs the principle of separation of concern to allocate different functionalities and responsibilities to different architectural elements comprising the system and it is easier to make changes to a software system that has a well-designed architecture. Conversely, bad architecture, manifested as architectural bad smells, can increase the complexity, possibly leading to poor software quality.

However, a software system's architecture is known to commonly undergo the phenomenon of architectural decay, where changes and design decisions are added to the system which may break the initially designed system's software architecture. Architectural decay has a negative impact on maintaining the system and results in defects and architectural problems in the system. Thus detecting and preferably avoiding decay will save considerable time and other resources from developers and stakeholders in a system. This dissertation targets empirical research in the domain of architecture-based software maintenance. It benefits from both fields of software architecture and mining software repository. The mining software repository (MSR) field investigates the rich data in source code repositories and defect repositories to uncover interesting information about software systems. For example data in source code repositories can be linked with data in defect repositories to observe what kind of changes would result in more defects. This would help to warn practitioners and developers about risky changes based on prior changes and faults.

In this research, I first investigate the impact of software architecture on defects from evolutionary history of software. To do that, I designed an empirical study to see whether there is a difference between types of changes made to a software system from software architecture perspective. Specifically I wanted to investigate the impact of co-changes involving several architectural modules versus co-changes localized within a single module. This provided empirical evidence for the importance of considering of software architecture while making changes to a system.

Next, I construct novel models that predict the quality of an architectural element by utilizing multiple architectural views (both structural and semantic) and architectural metrics as features for prediction. Using these models, I could predict low architectural quality, i.e., architectural decay in software systems. Engineers can significantly benefit from determining which architectural elements will decay before that decay actually occurs. Forecasting decay allows engineers to take steps to prevent decay, such as focusing maintenance resources on the architectural elements most likely to decay.

This research underlines the importance of software architecture in the construction and maintenance of software.