Numéro bibliothèque: ThèsedeDoctoratdel’universitéParisVI

The deployment of Self-Organizing Networks (SONs) raises a number of key design issues such as scalability, spontaneity, ease-of-management, adaptability, and dynamicallychanging topologies. In this context, routing is particularly challenging, and requires revisiting important components of the routing architecture/protocol suite, such as, establishing and managing an addressing scheme, providing a distributed location service, and defining a packet forwarding mechanism. In response to these requirements, Distributed Hash Tables (DHTs) were shown to be an efficient scheme for the design of large scale SONs. In this thesis, we envisage an adaptive addressing structure and a mathematical space that allow low management overhead and that ease DHT-based location in large scale SONs. We propose two systems to address these issues: Tribe and Twins. The mathematical architectures of Tribe and Twins implement DHT structures in two different ways. In Tribe, the structure is used for locating and routing, while in Twins, it is only used for locating. Tribe defines a tree-like mathematical structure to assign addresses to nodes, and to forward data, without any central control entity or positioning mechanism. The resulting structure allows performing location and forwarding by considering only neighboring information. Tribe has as advantages the simplicity of the addressing/routing structure construction, the small amount of information to implement routing, and the low signaling overhead. Twins specifies a multidimensional Cartesian addressing structure, which is a strict mathematical representation of the network geographic space. This representation is obtained through Hilbert space-filling curves. The geographic space is used for addressing and routing, while location is performed on the designed mathematical structure. When compared with the tree-like scheme, Twins improves the resilience and routing performance due to the flexibility in route selection. This thesis addresses the problem of efficiently routing content in large scale selforganizing networks. Both Tribe and Twins address fundamental issues in SONs and provide: (a) the use of well-defined mathematical spaces, which allow controlling overhead and complexity of management operations, and (b) scalable location services that are independent of spatial distribution of nodes over the topology.