Self-* and Predictability: are these conflicting system capabilities?

In principle, it would have been desirable for any system to have self-* capabilities when faced with changes that might occur either in the system or its environment. However, these capabilities are difficult to incorporate in certain classes of systems, mainly those in which uncertainties in their behaviours are not desirable. In this position paper, we discuss how the expected degree of autonomy of a system is related to the way in which a system can be described. The discussion is presented in the context of predictability and the assurances associated with the accuracy of that predictability. The subject of this paper are those systems known to be critical in the services they deliver, and which cannot incorporate high degree of uncertainty.