Thesis (M.S.)--Washington State University, August 2007. / Includes bibliographical references (p. 72-74).

Software-Intensive Cyber-Physical Systems (siCPS) are modular, open-ended, networked, large-scale embedded Information and Communication Technology (ICT) systems that are increasingly depending on software. They need to be both dependable and flexible to adapt to changes in their dynamic environments. This combination poses challenges in their design and development, as traditional model-driven design and development techniques cannot account for both dependability and self-adaptivity. The thesis proposes (1) a new, model-based design process for siCPS, which comprises both appropriate methods and models and deals with dependability and self-adaptivity, and (2) a mapping of the design models to implementation-level abstractions, which allows for model-driven development and early experimentations in siCPS. Specifically, the thesis delivers (1) by introducing and elaborating on the Invariant Refinement Method (IRM), and its extension for self-adaptivity, for the design of siCPS based on the ensemble paradigm. IRM was integrated into the ensemble development life cycle, a methodology for the development of autonomic ensemble-based systems. Contributing to (2), the thesis provides a mapping of the IRM concepts to the concepts of the DEECo components model. The mapping is supported by prototype...

Ph.D. / Information technology (IT) system development faces increasing challenges as a result of the complexity involved in the large number of interacting, distributed and concurrently executing components of systems. These components range from operating systems and virtual machines, through to the various frameworks, servers and libraries. To continue delivering on the current trend in resource requirements, hardware is increasingly parallel. The parallelisation of hardware indicates that software systems must be enabled to exploit these multicore, symmetric multiprocessing and distributed architectures as they become more mainstream. Parallelisation of IT systems adds to the number, distribution and concurrency of interacting components. In addition, dynamic self-optimising, selfhealing, self-configuring and self-protecting characteristics are required if systems are to continue operating effectively. The environment into which a system is eventually deployed is often either unknown or dynamic. An unknown environment is one where the exact details of resource availability, along with knowledge or control over concurrently executing systems, is not available beforehand. Added to this is a lack of foreknowledge surrounding the system’s environment, which may be dynamic, meaning it is likely to change during the system’s lifetime. Changes to the system’s environment include new infrastructure, different architectures, replacing old hardware and installing or upgrading software. The current approaches to overcoming unknown and dynamic environments tend to be top-down and centralised as is seen in the use of control theory by autonomic computing. There is, however, a growing realisation that centralised approaches add to the brittleness and complexity of the systems. What is needed is the self-adaptivity of an agent based approach, which is able to overcome these challenges relating to unknown and dynamic environments. Nature has dealt with the same challenges in a far more robust way by employing the principles of self-organising systems underpinning the control of complex adaptive systems. An example of nature’s solution is the self-organising system presented by the gene regulatory system coupled to cell fate and the cell cycle in multicellular organisms. Organisms are self-healing, self-protecting, self-optimising and self-configuring. They are also able, through ontogenesis, to self-adapt to their environments and grow to maximise their performance whilst still maintaining function.

The purpose of this project was to try to investigate one approach to the problem of automatically generating programs from some specification. Rather than following the approach which requires the user to define his problem using some formulation, it was decided to look at a class of problems that have similar solutions, but have many variations, and to try to design a system capable of obtaining user requirements and generating solutions tailored to these requirements. The aim was to design the system in such a way that it could be extended to cater for other classes of problems, so that eventually a system which could automatically generate program solutions for a range of problems might be developed. Intro. p. 1.

In recent years studies have been conducted which suggest that taking an architecture first approach to managing large software projects can reduce a significant amount of the uncertainty present in project estimates. As the project progresses, more concrete information is known about the planned system and less risk is present. However, the rate at which risk is alleviated varies across the life-cycle. Research suggests that there exists a significant drop off in risk when the architecture is developed. Software risk assessment techniques have been developed which attempt to quantify the amount of risk that varying uncertainties convey to a software project. These techniques can be applied to architecture specific issues to show that in many cases, conducting an architecture centric approach to development will remove more risk than the cost of developing the architecture. By committing to developing the architecture prior to the formal estimation process, specific risks can be more tightly bounded, or even removed from the project. The premise presented here is that through the process of architecture-centric management, it is possible to remove substantial risk from the project. This decrease in risk exceeds that at other phases of the life-cycle, especially in comparison of the effort involved. Notably, at architecture, a sufficient amount knowledge is gained by which effort estimations may be tightly bounded, yet the project is early enough in the life-cycle for proper planning and scheduling. Thus, risk is mitigated through the increase in knowledge and the ability to maintain options at an early point. Further, architecture development and evaluation has been shown to incorporate quality factors normally insufficiently considered in the system design. The approach taken here is to consider specific knowledge gained through the architecting process and how this is reflected in parametric effort estimation models. This added knowledge is directly reflected in risk reduction. Drawing on experience of architecture researchers as well as project managers employing this approach, this thesis considers what benefits to the software development process are gained by taking this approach. Noting a strong reluctance of owners to incorporate solid software engineering practices, the thesis concludes with an outline for an experiment which goes about proving the reduction in risk at architecture exceeds the cost of that development. / Master of Science

Software developers must produce software that can be changed without the risk of degrading the software architecture. One way to address software changes is to classify their causes and effects. A software change classification mechanism allows engineers to develop a common approach for handling changes. This information can be used to show the potential impact of the change. The goal of this research is to develop a change classification scheme that can be used to address causes of architectural degradation. This scheme can be used to model the effects of changes to software architecture. This research also presents a study of the initial architecture change classification scheme. The results of the study indicated that the classification scheme was easy to use and provided some benefit to developers. In addition, the results provided some evidence that changes of different types (in this classification scheme) required different amounts of effort to implement.

A majority of software development today involves maintenance or evolution of legacy systems. Evolving these legacy systems, while maintaining good software design principles, is a significant challenge. Research has shown the benefits of using software architecture as an abstraction to analyze quality attributes of proposed designs. Unfortunately, for most legacy systems, a documented software architecture does not exist. Developing a good architectural description frequently requires extensive experience on the part of the developer trying to recover the legacy system's architecture. This work first describes a four-phase process that provides a framework within which architectural recovery activities can be automated. These phases consist of: extraction (obtaining a subset of information about the legacy system from a single source), classification (partitioning the information based upon its viewpoint), union (combining all the information in a particular viewpoint into a candidate view), and fusion (cross-checking all candidate views for consistency. The work then concentrates on the major problem facing automated architectural recovery---the concept assignment problem. To overcome this problem, a technique called semantic approximation is presented and validated via experimental results. Semantic approximation uses a combination of text data mining and a mathematical technique called concept analysis to build a lattice of similar concepts between higher-level domain information and low-level code concepts. The experimental data reveals that while semantic approximation does improve results over the more traditional lexical and topological approaches, it does not yet fully solve the concept assignment problem.

Arquitetura de software desempenha um importante papel no desenvolvimento de software, quando explicitamente documentada, ela melhora o entendimento sobre o sistema implementado e torna possível entender a forma com que requisitos não funcionais são tratados. Apesar da relevância da arquitetura de software, muitos sistemas não possuem uma arquitetura documentada, e nos casos em que a arquitetura existe, ela pode estar desatualizada por causa da evolução descontrolada do software. O processo de recuperação de arquitetura de um sistema depende principalmente do conhecimento que as pessoas envolvidas com o software tem. Isso acontece porque a recuperação de arquitetura é uma tarefa que demanda muita investigação manual do código fonte. Pesquisas sobre recuperação de arquitetura objetivam auxiliar esse processo. A maioria dos métodos de recuperação existentes são baseados em dependência entre elementos da arquitetura, padrões arquiteturais ou similaridade semântica do código fonte. Embora as abordagem atuais ajudem na identi cação de módulos arquiteturais, os resultados devem ser melhorados de forma signi cativa para serem considerados con áveis. Então, nesta dissertação, objetivamos melhorar o suporte a recuperação de arquitetura explorando diferentes fontes de informação e técnicas de aprendizado de máquina. Nosso trabalho consiste de uma análise, considerando cinco estudo de casos, da utilidade de usar um conjunto de características de código (features, no contexto de aprendizado de máquina) para agrupar elementos em módulos da arquitetura. Atualmente não são conhecidas as características que afetam a identificação de papéis na arquitetura de software. Por isso, nós avaliamos a relação entre diferentes conjuntos de características e a acurácia obtida por um algoritmo não supervisionado na identificação de módulos da arquitetura. Consequentemente, nós entendemos quais dessas características revelam informação sobre a organização de papéis do código fonte. Nossa abordagem usando características de elementos de software atingiu uma acurácia média significativa. Indicando a relevância das informações selecionadas para recuperar a arquitetura. Além disso, nós desenvolvemos uma ferramenta para auxílio ao processo de recuperação de arquitetura de software. Nossa ferramenta tem como principais funções a avaliação da recuperação de arquitetura e apresentação de diferentes visualizações arquiteturais. Para isso, apresentamos comparações entre a arquitetura concreta e a arquitetura sugerida. / Software architecture plays an important role in the software development, and when explicitly documented, it allows understanding an implemented system and reasoning about how non-functional requirements are addressed. In spite of that, many developed systems lack proper architecture documentation, and if it exists, it may be outdated due to software evolution. The process of recovering the architecture of a system depends mainly on developers' knowledge requiring a manual inspection of the source code. Research on architecture recovery provides support to this process. Most of the existing approaches are based on architectural elements dependency, architectural patterns or source code semantics, but even though they help identifying architectural modules, the obtained results must be signi cantly improved to be considered reliable. We thus aim to support this task by the exploitation of di erent code-oriented information and machine learning techniques. Our work consists of an analysis, involving ve case studies, of the usefulness of adopting a set of code-level characteristics (or features, in the machine learning terminology) to group elements into architectural modules. The characteristics mainly source code metrics that a ect the identi cation of what role software elements play in software architecture are unknown. Then, we evaluate the relationship between di erent sets of characteristics and the accuracy achieved by an unsupervised algorithm the Expectation Maximization in identifying architectural modules. Consequently, we are able to understand which of those characteristics reveal information about the source code structure. By the use of code-oriented information, our approach achieves a signi cant average accuracy, which indicates the importance of the selected information to recover software architecture. Additionally, we provide a tool to support research on architecture recovery providing software architecture measurements and visualizations. It presents comparisons between predicted architectures and concrete architectures.

Arquitetura de software desempenha um importante papel no desenvolvimento de software, quando explicitamente documentada, ela melhora o entendimento sobre o sistema implementado e torna possível entender a forma com que requisitos não funcionais são tratados. Apesar da relevância da arquitetura de software, muitos sistemas não possuem uma arquitetura documentada, e nos casos em que a arquitetura existe, ela pode estar desatualizada por causa da evolução descontrolada do software. O processo de recuperação de arquitetura de um sistema depende principalmente do conhecimento que as pessoas envolvidas com o software tem. Isso acontece porque a recuperação de arquitetura é uma tarefa que demanda muita investigação manual do código fonte. Pesquisas sobre recuperação de arquitetura objetivam auxiliar esse processo. A maioria dos métodos de recuperação existentes são baseados em dependência entre elementos da arquitetura, padrões arquiteturais ou similaridade semântica do código fonte. Embora as abordagem atuais ajudem na identi cação de módulos arquiteturais, os resultados devem ser melhorados de forma signi cativa para serem considerados con áveis. Então, nesta dissertação, objetivamos melhorar o suporte a recuperação de arquitetura explorando diferentes fontes de informação e técnicas de aprendizado de máquina. Nosso trabalho consiste de uma análise, considerando cinco estudo de casos, da utilidade de usar um conjunto de características de código (features, no contexto de aprendizado de máquina) para agrupar elementos em módulos da arquitetura. Atualmente não são conhecidas as características que afetam a identificação de papéis na arquitetura de software. Por isso, nós avaliamos a relação entre diferentes conjuntos de características e a acurácia obtida por um algoritmo não supervisionado na identificação de módulos da arquitetura. Consequentemente, nós entendemos quais dessas características revelam informação sobre a organização de papéis do código fonte. Nossa abordagem usando características de elementos de software atingiu uma acurácia média significativa. Indicando a relevância das informações selecionadas para recuperar a arquitetura. Além disso, nós desenvolvemos uma ferramenta para auxílio ao processo de recuperação de arquitetura de software. Nossa ferramenta tem como principais funções a avaliação da recuperação de arquitetura e apresentação de diferentes visualizações arquiteturais. Para isso, apresentamos comparações entre a arquitetura concreta e a arquitetura sugerida. / Software architecture plays an important role in the software development, and when explicitly documented, it allows understanding an implemented system and reasoning about how non-functional requirements are addressed. In spite of that, many developed systems lack proper architecture documentation, and if it exists, it may be outdated due to software evolution. The process of recovering the architecture of a system depends mainly on developers' knowledge requiring a manual inspection of the source code. Research on architecture recovery provides support to this process. Most of the existing approaches are based on architectural elements dependency, architectural patterns or source code semantics, but even though they help identifying architectural modules, the obtained results must be signi cantly improved to be considered reliable. We thus aim to support this task by the exploitation of di erent code-oriented information and machine learning techniques. Our work consists of an analysis, involving ve case studies, of the usefulness of adopting a set of code-level characteristics (or features, in the machine learning terminology) to group elements into architectural modules. The characteristics mainly source code metrics that a ect the identi cation of what role software elements play in software architecture are unknown. Then, we evaluate the relationship between di erent sets of characteristics and the accuracy achieved by an unsupervised algorithm the Expectation Maximization in identifying architectural modules. Consequently, we are able to understand which of those characteristics reveal information about the source code structure. By the use of code-oriented information, our approach achieves a signi cant average accuracy, which indicates the importance of the selected information to recover software architecture. Additionally, we provide a tool to support research on architecture recovery providing software architecture measurements and visualizations. It presents comparisons between predicted architectures and concrete architectures.

Arquitetura de software desempenha um importante papel no desenvolvimento de software, quando explicitamente documentada, ela melhora o entendimento sobre o sistema implementado e torna possível entender a forma com que requisitos não funcionais são tratados. Apesar da relevância da arquitetura de software, muitos sistemas não possuem uma arquitetura documentada, e nos casos em que a arquitetura existe, ela pode estar desatualizada por causa da evolução descontrolada do software. O processo de recuperação de arquitetura de um sistema depende principalmente do conhecimento que as pessoas envolvidas com o software tem. Isso acontece porque a recuperação de arquitetura é uma tarefa que demanda muita investigação manual do código fonte. Pesquisas sobre recuperação de arquitetura objetivam auxiliar esse processo. A maioria dos métodos de recuperação existentes são baseados em dependência entre elementos da arquitetura, padrões arquiteturais ou similaridade semântica do código fonte. Embora as abordagem atuais ajudem na identi cação de módulos arquiteturais, os resultados devem ser melhorados de forma signi cativa para serem considerados con áveis. Então, nesta dissertação, objetivamos melhorar o suporte a recuperação de arquitetura explorando diferentes fontes de informação e técnicas de aprendizado de máquina. Nosso trabalho consiste de uma análise, considerando cinco estudo de casos, da utilidade de usar um conjunto de características de código (features, no contexto de aprendizado de máquina) para agrupar elementos em módulos da arquitetura. Atualmente não são conhecidas as características que afetam a identificação de papéis na arquitetura de software. Por isso, nós avaliamos a relação entre diferentes conjuntos de características e a acurácia obtida por um algoritmo não supervisionado na identificação de módulos da arquitetura. Consequentemente, nós entendemos quais dessas características revelam informação sobre a organização de papéis do código fonte. Nossa abordagem usando características de elementos de software atingiu uma acurácia média significativa. Indicando a relevância das informações selecionadas para recuperar a arquitetura. Além disso, nós desenvolvemos uma ferramenta para auxílio ao processo de recuperação de arquitetura de software. Nossa ferramenta tem como principais funções a avaliação da recuperação de arquitetura e apresentação de diferentes visualizações arquiteturais. Para isso, apresentamos comparações entre a arquitetura concreta e a arquitetura sugerida. / Software architecture plays an important role in the software development, and when explicitly documented, it allows understanding an implemented system and reasoning about how non-functional requirements are addressed. In spite of that, many developed systems lack proper architecture documentation, and if it exists, it may be outdated due to software evolution. The process of recovering the architecture of a system depends mainly on developers' knowledge requiring a manual inspection of the source code. Research on architecture recovery provides support to this process. Most of the existing approaches are based on architectural elements dependency, architectural patterns or source code semantics, but even though they help identifying architectural modules, the obtained results must be signi cantly improved to be considered reliable. We thus aim to support this task by the exploitation of di erent code-oriented information and machine learning techniques. Our work consists of an analysis, involving ve case studies, of the usefulness of adopting a set of code-level characteristics (or features, in the machine learning terminology) to group elements into architectural modules. The characteristics mainly source code metrics that a ect the identi cation of what role software elements play in software architecture are unknown. Then, we evaluate the relationship between di erent sets of characteristics and the accuracy achieved by an unsupervised algorithm the Expectation Maximization in identifying architectural modules. Consequently, we are able to understand which of those characteristics reveal information about the source code structure. By the use of code-oriented information, our approach achieves a signi cant average accuracy, which indicates the importance of the selected information to recover software architecture. Additionally, we provide a tool to support research on architecture recovery providing software architecture measurements and visualizations. It presents comparisons between predicted architectures and concrete architectures.