"Adaptive Dependability in Distributed Systems";
TU Wien, Institut für Informationssysteme, AB Verteilte Systeme,
While dependability becomes a cornerstone of the information society, it is impaired by change, imprecision, and emerging behavior due to scale, dynamism, and heterogeneity. To address these challenges, dependability has to be persisted when facing changes, which is called resilience. In this thesis, adaptivity is envisaged to achieve resilience in this thesis.
The key contribution of this thesis is to propose open or closed loop control to adaptively modify the strength of coupling between the system´s constituents as the most promising approach to balance its overall dependability and security properties.
This idea is elaborated in the form of three key principles that apply the well-known concept of separation of concerns to the run-time of software: Balancing and middleware support, (temporal) decoupling, and run-time processable requirements.
Based on a multitude of industrial applications and prototypes in the fields of safety-critical, resource-critical, and mission-critical systems, the three key principles are validated and quantitatively evaluated for three scenarios: Distributed object systems, service-oriented systems, and web systems with high security demands.
Results show, resilience can indeed be improved with the proposed approach, but may come at the cost of additional complexity. Depending on architectural style and used technology, this may require additional tool support.
Created from the Publication Database of the Vienna University of Technology.