Building autonomic components: The SelfLets approach

Autonomic computing is an emergent field aiming at the development of large-scale, self-managing, distributed component-based systems. This paper presents the model and the architecture of an autonomic computing element called SelfLet, which is a building component that can be used to create autonomic systems. SelfLets can be defined by specifying their goal, behaviors, services they need to use and/or provide, and autonomic policies guiding their self-management. The SelfLet architecture has been implemented in Java and offers programming abstractions suitable to implement an application-specific logic as well as autonomic policies. As a case study we have implemented a pervasive autonomic system that manages electrical power balancing in intelligent cooperating buildings.