Overlay data structures are a powerful mechanism to provide application components with context-information and to let them interact in dynamic-network scenarios like mobile ad-hoc networks and pervasive computing. These overlays can be propagated across a network in order to support components’ context awareness and coordination activities. We present a modeling framework and some autonomic algorithms to create overlay data structures that are able to self-maintain their intended distribution under a number of circumstances. The paper presents some experiments and performance measures to validate our approach and to show its scalability.