A framework for dependable adaptation in probabilistic environments

Distributed applications executing in probabilistic environments, like the Internet, often need to make timing assumptions, for instance about the maximum message delay or the round-trip delay. In the case of adaptive systems these bounds should be computed at run-time, using probabilistic or other more or less ad hoc approaches, typically with the objective of improving the application performance. From a dependability perspective, however, the concern is to secure some properties on which the application can rely. Building on previous work where we set the stage for dependable adaptation, we introduce a general framework for implementing improved solutions for dependable adaptation. During its lifetime, a system alternates periods where its temporal behavior is well characterized, with transition periods where a variation of the environment conditions occurs. The proposed framework rests on the detection of these stable and transient phases, and it is generic in the sense that different detection mechanisms can be included. We use a set of simple phase detection mechanisms inspired in well-known probabilistic tests, along with synthetically generated data flows, to illustrate how the framework can be applie