Project "Adaptation in Autonomic Comuting"
Autonomic Computing is a paradigm first called so by J.O. Khepart and D.M. Chess of IBM research. It sums up a set of system attributes like "self-healing", "self-organizing", "self-configuring" and "self-protection" which in sum aim at systems that are able to administer themselves along the directives of a policy given by an administrator. The background is the observation that administration tasks become more and more complex and administrators are occupied by dealing with details rather than concentrating on the overall goals of administration. Therefore the systems should be able to assist administrators by making certain administration tasks automatically like the autonomic nervous system does for animals and humans. Our thesis is that context adaptation and reconfiguration fit very well as Techniques to support the former goals.
The objectives of the first project phase (AAC) in detail are:
Objective 1: requirements for development process- State of the art analysis
- Overview of methodology
- Big picture of RE, design and implementation - classification of current approaches
- elicitation of requirements for autonomic computing application
- Modeling domain comprehensive automation (e.g. resolving requirements conflicts)
- Enabling better quality of recovery compared to static prioritization
- Transition to a logical/functional architecture
- Big picture => Step-over from functional to technical architecture
- Identification of patterns, reference architecture, building blocks etc.
- Validation of concepts by means of prototypical experiments
- Estimation of performance and comparison
The second project phase (AAC2):
AAC2 takes the results from AAC and geeralizes them. Now it is of interest how the concepts can be applied if other frameworks than CAWAR are used. Therefore the conceps have to be discussed in more general and it shold be worked out which aspects a framework must satisfy to allow to apply the concepts.
The notion of service is discussed. E.g. services could be regarded as stateless or statefull. In either case other aspects have to be considered while applying the general concepts. The project discusses a broad variety of service notions and alaylzes their appropriateness to adaptation and the modes of application e.g. how are services started and stopt etc.
A second objective is the revision and generalisation of the proposed design heuristics. The heuristics are now abstracted in a way that it is possible to apply them to other frameworks.
The question about the extent of adaptation is discussed. As context adaptive systems are reactive systems in general that model certain aspects in a notable manner the extent of this modelling is a design decision. It is important to give hints about the goals of modeling as adaptation and the pros and cons about it.
Finally to demonstrate the feasability and advances of adaptation and reconfiguration in this domain we create a demonstrator in the domains of health care and facility management.