An adaptive distributed system based on conditional dependencies

Distributed programming is applicable in a wide range of domains, such as control systems. These systems are subject to mutable environments and must also provide a time-varying adaptive set of functionalities. An adaptive system is defined by means of two main features: (i) an approach to specify the correspondence between objects and resources, and (ii) the related mechanisms to dynamically assure that the correct set of objects is running. The conditional dependency concept has proven to be an efficient way to specify this kind of correspondence between resources and objects for adaptive systems in the context of multimedia distributed processing (in the ServiMidia Project). This paper presents a framework for designing general-purpose adaptive systems anchored on the concept of conditional dependencies. As the use of such an approach may lead to inconsistent system behaviour when specifying very large systems, we show that it is easy to design verification tools based on the same framework mechanisms.