Product-Focused Software Process Improvement. Industry-, Workshop-, and Doctoral Symposium Papers

As the landscape of cloud computing continues to transform at a rapid pace, organizations seek to optimize operational costs, enhance system flexibility, and adjust to changing market demands. This paper reports a case study within a Finnish startup operating in self-service mobile payment systems. It provides a detailed analysis of migrating an application architecture from DigitalOcean to AWS (Amazon Web Services) using the AWS Cloud Development Kit (CDK), providing insights into the strategies, challenges, and outcomes of the process. The key findings suggest that while cloud migration involves considerable technical and operational challenges, the utilization of IaC (Infrastructure as Code) can significantly mitigate these obstacles, enhancing the speed, reliability, and security of the migration process. This research contributes to the limited literature on cloud-to-cloud migration, highlighting the effective use of IaC to facilitate transitions between cloud environments.

In this case study we compare six different metrics and their suitability for productivity benchmarking on development output level. We use detailed data from four software industry projects performed by one company with overall 264 months of development and 1.1 million source lines of code. Code change, absolute growth and number of commits as well as invested effort are measured in consecutive 3-month periods. This allows us to observe alterations in productivity throughout the course of a project as well as inter-project comparisons. We find correlations between effort and the chosen output metrics as well as significant and explainable productivity differences between projects and project phases.

We also analyze whether the use of a clone detection algorithm can improve measurement by adjusting for copy & paste additions and renamed or moved code, and find that a small benefit exists. The redundancy-adjusted amount of code tokens added or modified seems to be the best metric among the selected, in particular in ongoing development where an already existing codebase is changed. Number of commits and absolute growth may complement the picture.

Autonomous driving software generates enormous amounts of data every second, which software development organizations save for future analysis and testing in the form of logs. However, given the vast size of this data, locating specific scenarios within a collection of vehicle logs can be challenging. Writing the correct SQL queries to find these scenarios requires engineers to have a strong background in SQL and the specific databases in question, further complicating the search process. This paper presents and evaluates a pipeline that allows searching for specific scenarios in log collections using natural language descriptions instead of SQL. The generated descriptions were evaluated by engineers working with vehicle logs at the Zenseact on a scale from 1 to 5. Our approach achieved a mean score of 3.3, demonstrating the potential of using a multi-model architecture to improve the software development workflow. We also present an interface that can visualize the query process and visualize the results.

Generative AI is transforming software development, particularly in unit and regression testing. However, it’s rarely used in Hardware-in-the-Loop (HIL) testing due to hardware-specific environments. This paper examines integrating GitHub Copilot into automotive test automation frameworks, focusing on Volvo’s Test Automation Framework (TAF). It explores how Copilot can automate test case generation and compares AI-generated test cases with manually written ones in terms of reliability and robustness. Using an iterative action research methodology, the study evaluates the functional suitability of AI-generated test cases and the challenges of integration. Results show that in the first iteration, 23% of AI-generated test cases passed in Jenkins and received high functionality scores. In the second iteration, this increased to 36%. These findings highlight the potential of Generative AI to enhance HIL testing.

The AI Act’s (AIA) requirements for high-risk AI systems affect many aspects of modern software systems. Knowing which AIA-related technical challenges are relevant to different companies is essential to focus compliance-oriented research on the aspects that matter. We therefore conducted an interview study in collaboration with a case company that specializes in network video solutions within the security and surveillance industry. External experts enrich the study for a broader industry perspective. The goal was to analyze the case company’s readiness for the AIA’s high-risk requirements, based on methods and techniques already established prior to the legislation. Our results yielded a positive sentiment towards the regulation and the planning security that it brings, although a high workload was expected. We identified a solid foundation with well-established practices to build upon for the requirements on cybersecurity, human oversight, record-keeping, and technical documentation. However, we also report several open challenges, mainly connected to the requirement on data quality and governance, followed by accuracy, robustness, and cybersecurity. The AIA specifically demands a post-market monitoring system (Art 72) and the right to an explanation of individual decision-making (Art 86). These two obligations were identified as especially challenging by the respondents. The result of this study is expected to steer future compliance-oriented work toward pressing challenges.

Continuous deployment is a significant trend in software development, yet its adoption and potential benefits within the public sector remain under-researched. This paper examines a case study of continuous deployment implementation in a public sector project undertaken by Solita, a software development company, for a client utilizing agile methodologies. The study provides a comprehensive overview of the motivations, benefits, and challenges encountered during continuous deployment adoption. This study contributes to the growing body of knowledge on continuous deployment by providing valuable insights into its application within the public sector context, offering practical recommendations for overcoming challenges and achieving successful implementation.

Architectural Technical Debt (ATD) is a persistent challenge in software development, often hindered by the absence of comprehensive architectural documentation. This research presents an automated approach to detect source code files indicative of ATD within a version control system. By analyzing code metrics, change history, and architectural smells, we identify files exhibiting signs of increasing complexity and maintenance effort. Our method was applied to the Apache Cassandra repository and validated through interviews with Ericsson developers. Results indicate a strong correlation between files with specific architectural smells, frequent modifications, and growing complexity, and the presence of ATD. This study demonstrates the feasibility of using source code analysis to systematically identify potential architectural issues, aiding developers in prioritizing refactoring efforts and improving software quality.

In domains such as aerospace, automotive, and medical devices, high-quality software is key due to the critical nature of these applications and the low margin for failure. Ensuring the quality of software is essential to prevent potentially catastrophic outcomes. In this paper, we present an approach to derive quality assurance plans from well-defined process models. Utilizing the GQM model, we derive quality requirements and metrics based on a process example, which we realize in the process management tool Stages. Based on the exemplary realization, we simulate 100 projects to provide data used in a recommendation system that enriches the process model such that quality assurance plans including the metrics of interest can be generated from the process model. Our findings show that, given sufficient data is available, context-specific quality assurance plans can be generated and which particular steps have to be taken to realize the overall concept in the studied tool.

Many maintenance and evolution tasks in software engineering depend on the availability of logical code segments, especially AI- and data-driven approaches rely on segmented source code. In order to obtain logical segments a manual step is typically necessary. However, manual segmentation requires a basic understanding of the source code and delays the application of code analysis and refactoring tools. Automatic code segmentation provides an efficient way of extracting code snippets for further analysis to provide developers with actionable insights on software products and processes. Rule-based approaches rely on syntactic boundaries and lack the applicability of segmentation on multiple languages. In this article, we present our approach to learning logical code snippets using a BiLSTM neural network model. Driven by the requirements of an industrial use case, we train two models, one on the programming languages Go, Java, JavaScript, Python, PHP, and Ruby, the other model is additionally trained on ABAP (“Advanced Business Application Programming”) snippets. To evaluate the performance of the models, we use real-world samples used in the SAP applications of our industry partner. We also compare the predictions of the model, with an accuracy >98%, to the results of human experts for segmenting ABAP code to evaluate whether AI-based code segmentation is perceived as effective by practitioners in our industrial use case. The study shows that only 42–51% of the predicted ABAP code snippets match the manual segmentation of the experts.

In this paper, we evaluate the feasibility of quantum algorithms for practical applications and categorize them into three types: green, yellow, and red. Green means the most feasible, while red means the least feasible. We select four quantum algorithms from the Algebraic and Number Theoretic fields, four from the Optimization field, six from the Machine Learning field, and four from the Oracular field to assess their feasibility. Our results show that some quantum algorithms can be applied to solve real-life problems in the near future, while other fields may take a long time to be practically applied. The feasibility assessment is obtained by considering whether there are requirements in the quantum algorithms that are hard to satisfy with current quantum technologies and predicting how much time it will require for those requirements to be satisfied in the future. We also provide a table summarizing the feasibility evaluation results of these quantum algorithms.

Identifying possible use cases for quantum computers is important to evaluate the potential. This work-in-progress paper explores their potential for addressing the short-term generation scheduling (STGS) problem in hydropower plants. By working towards reformulating the STGS problem as a quadratic unconstrained binary optimization (QUBO) problem, we aim to leverage the capabilities of quantum computers to find optimal solutions. Initial results using piecewise linear approximation indicate promising outcomes. Further research will focus on the QUBO formulation, implementing the approach on quantum hardware, and assessing its performance.

Testing of applications that use data from e-Government services as input requires test data that is real-life like but where the privacy of personal information is guaranteed. Many approaches exist for creating high-quality synthetic test data, but most of them need real-life raw data as input. Our research aims to develop and evaluate an approach for generating and evolving real-life like synthetic test data without using real-world raw data. The expected benefit of our research is to enable the creation and evolvement of meaningful and real-life like synthetic test data in situations where real-life raw data is not accessible due to privacy reasons.

The growing complexity, the ever-ending demands for new features, and the need to become faster to remain competitive force software development organizations to rethink their development and value delivery practices. While continuous delivery has become more popular, it still relies mainly on internal metrics, ad-hoc data, and expert opinions. As a result, software organizations stumble to find the balance between improving internal system quality and delivering external value. In fact, understanding and measuring customer value is on itself essential. In this PhD project, we aim for a better understanding of customer value and develop measurement instruments to be integrated with internal perspectives to drive proactive and continuous internal improvement while delivering relevant customer value.