Global Observation for Distributed Autonomic Networks

While much is known regarding software engineering practices for dependable software systems; the extreme scale, complexity and dynamics of modern software has pushed conventional software engineering tools and techniques to their acceptable limits. Consequently, over the last decade, this has accelerated research into non-conventional methods, many of which are inspired by social and/or biological systems models. In addition, existing federated approaches to distributed computation and control, such as Multi-Agent-Systems fail to satisfactorily address how global control may be enacted upon the whole system and how an individual may take on specified monitoring duties - although methods of interaction between federated individuals is well understood. Hence, this research sets out to study the engineering concerns for observation/monitoring of large-scale networks of autonomic systems. As such, this research examines methods that can be used to manage scale; looks to generalise and formalise them within a software engineering approach that guides the development of an automated adaptive observation subsystem - the Global Observer Model. This approach uses a model-based representation of the observed system, represented by appropriately attached modelled elements; effectively adapters between the underlying system and the observation subsystem. The approach is tested by a case study with further specific applications outlined for ongoing and future development.