Safeware Engineering (United States)

Architecture erosion occurs when a software system's implemented architecture diverges from the intended architecture over time. Studies show erosion impacts development, maintenance, and evolution since it accumulates imperceptibly. Identifying early symptoms like architectural smells enables managing erosion through refactoring. However, research lacks comprehensive understanding of erosion, unclear which symptoms are most common, and lacks detection methods. This thesis establishes an erosion landscape, investigates symptoms, and proposes identification approaches. A mapping study covers erosion definitions, symptoms, causes, and consequences. Key findings: 1) "Architecture erosion" is the most used term, with four perspectives on definitions and respective symptom types. 2) Technical and non-technical reasons contribute to erosion, negatively impacting quality attributes. Practitioners can advocate addressing erosion to prevent failures. 3) Detection and correction approaches are categorized, with consistency and evolution-based approaches commonly mentioned.An empirical study explores practitioner perspectives through communities, surveys, and interviews. Findings reveal associated practices like code review and tools identify symptoms, while collected measures address erosion during implementation. Studying code review comments analyzes erosion in practice. One study reveals architectural violations, duplicate functionality, and cyclic dependencies are most frequent. Symptoms decreased over time, indicating increased stability. Most were addressed after review. A second study explores violation symptoms in four projects, identifying 10 categories. Refactoring and removing code address most violations, while some are disregarded.Machine learning classifiers using pre-trained word embeddings identify violation symptoms from code reviews. Key findings: 1) SVM with word2vec achieved highest performance. 2) fastText embeddings worked well. 3) 200-dimensional embeddings outperformed 100/300-dimensional. 4) Ensemble classifier improved performance. 5) Practitioners found results valuable, confirming potential.An automated recommendation system identifies qualified reviewers for violations using similarity detection on file paths and comments. Experiments show common methods perform well, outperforming a baseline approach. Sampling techniques impact recommendation performance.

Engineering of embedded systems is considered highly complex, due to the need for integrating multi-site, multi-lifecycle, multi-disciplinary, and multi-organization approaches. However, such challenges have not been comprehensively addressed in existing engineering processes. To this end, this study proposes a process framework to tackle these challenges, focusing on its meta-model, i.e. A set of elements used for process instantiation. As an exploratory validation of the proposed meta-model we conducted two focus groups with experts on embedded systems, resulting in positive feedback and industrial acceptance.

Refactoring, aims to improve the design of existing code to cope with foreseen software architecture evolution. The selection of the optimum refactoring strategy can be a daunting task involving the identification of refactoring candidates, the determination of which refactorings to apply and the assessment of the refactoring impact on software product quality characteristics. As such, the benefits from refactorings are measured from the quality advancements achieved through the application of state of the art structural quality assessments on refactored code. Perceiving refactoring trough the lens of value creation, the optimum strategy should be the one that maximizes the endurance of the architecture in future imposed changes. We argue that an alternative measurement and examination of the refactoring success is possible, one, that focuses on the balance between effort spent and anticipated cost minimization. In this arena, traditional, quality evaluation methods fall short in examining the financial implications of uncertainties imposed by the frequent updates/modifications and by the dynamics of the XP programming. In this paper we apply simple Real Options Analysis techniques and we perceive the selection of the optimum refactoring strategy as an option capable of generating value (cost minimization) upon adoption. Doing so, we link the endurance of the refactored architecture to its true monetary value. To get an estimation of the expected cost that is needed to apply the considered refactorings and to the effect of applying them, in the cost of future adoptions we conducted a case study. The results of the case study suggest that every refactoring can be associated with different benefit levels during system extension.

This decision support system was built to assist project managers to design and tailor the software development process to their specific projects for achieving required software reliability values. The decision support system has two subsystems: a knowledge-based expert system for reliability engineering strategy selection and a process simulator for strategy assessment. We present here the knowledge-based component for strategy selection assistance. The paper describes the framework for identifying software reliability engineering practices and the decision factors for selecting a subset of practices appropriate for the project, taking into account various project constraints. After identifying heuristics for making the selection decisions, we have developed a knowledge base and implemented an expert system to help a less experienced project manager with engineering the process in a product quality driven manner. The expert system is a rule based fuzzy logic system, developed by using the off-the-shelf FuzzyCLIPS shell.

Eine der grundlegenden Fähigkeiten von Digitalrechnern ist, Informationen speichern zu können. Der Benutzer kann sich gespeicherte Informationen mitteilen lassen. Er kann Speicherinhalte vorgeben und