by
Abstract:
Self-adaptation is an effective approach to overcome the complexity of dynamically changing requirements and application domains. Systems augmented with this capability are able to modify themselves in order to satisfy certain objectives and goals without the need of human interference. A key requirement of self-adaptive systems is dependability. This requirement has a strong relationship with goals of self-healing and self-optimization and covers aspects like availability, reliability, maintainability or safety. A non-trivial task is to design and implement dependable software systems. Several approaches already tackled this challenge by adopting findings of the field of mathematical verification theory, artificial intelligence or software design. However, it is still the focus of current research efforts. In this thesis, we present another AI-based approach, that utilizes a system of constraints in order to cope with this demanding challenge. The thesis aims to provide the existing modular Rice Bidding system (mRUBiS) with a level of dependability by augmenting it with self-healing and self-optimization capabilities. The optimization and healing processes are based on a set of redeployments of software services to dedicated hardware. Our research focuses on the establishment of a hardware model for simulating deployment scenarios. Furthermore, it copes with the design and the implementation of a constraintbased planning approach. We outline different strategies, that optimize and accelerate this approach and study their effectiveness as valuable part of our system. We evaluate our implementation based on a set of manually created scenarios. They demonstrate, that the mRUBiS self-adaptive system is able to perform deployment modifications proven to be optimal. At the end, we discuss the practicability of our constraintbased approach and reason about scalability concerns.
Reference:
Constraint-based Self-optimisation of Deployments for modular Systems, (), Master's thesis, Hasso-Plattner-Institut für Digital Engineering, Universität Potsdam, 2018.
Bibtex Entry:
@MastersThesis{Naumann18,
AUTHOR = {Naumann, Tim},
TITLE = {{Constraint-based Self-optimisation of Deployments for modular Systems, }},
YEAR = {2018},
MONTH = {May},
SCHOOL = {Hasso-Plattner-Institut für Digital Engineering, Universität Potsdam},
PDF = {uploads/pdf/Naumann18.pdf},
OPTacc_pdf = {},
ABSTRACT = {Self-adaptation is an effective approach to overcome the complexity of dynamically changing requirements and application domains. Systems augmented with this capability are able to modify themselves in order to satisfy certain objectives and goals without the need of human interference. A key requirement of self-adaptive systems is dependability. This requirement has a strong relationship with goals of self-healing and self-optimization and covers aspects like availability, reliability, maintainability or safety. A non-trivial task is to design and implement dependable software systems. Several approaches already tackled this challenge by adopting findings of the field of mathematical verification theory, artificial intelligence or software design. However, it is still the focus of current research efforts. In this thesis, we present another AI-based approach, that utilizes a system of constraints in order to cope with this demanding challenge. The thesis aims to provide the existing modular Rice Bidding system (mRUBiS) with a level of dependability by augmenting it with self-healing and self-optimization capabilities. The optimization and healing processes are based on a set of redeployments of software services to dedicated hardware. Our research focuses on the establishment of a hardware model for simulating deployment scenarios. Furthermore, it copes with the design and the implementation of a constraintbased planning approach. We outline different strategies, that optimize and accelerate this approach and study their effectiveness as valuable part of our system. We evaluate our implementation based on a set of manually created scenarios. They demonstrate, that the mRUBiS self-adaptive system is able to perform deployment modifications proven to be optimal. At the end, we discuss the practicability of our constraintbased approach and reason about scalability concerns.}
}