Thèse de doctorat en Informatique
Sous la direction de Djamal Zeghlache.
Soutenue en 2007
Pas de résumé disponible.
A P2P-based naming system for service discovery and networking in wide-area dynamic environments
The world is becoming one large pervasive environment where devices and services autonomously work together, without human intervention. These devices and services are deployed in Clusters, forming dynamic communities of interacting entities. To link clients in any Cluster with remote services in other Clusters, devices and nodes have to capable of remotely discovering services throughout the Clusters. Wide-area service discovery has emerged as a potential solution for achieving the requirements of these pervasive environments and services. In highly dynamic environments, applications interact with frequently changing services and achieving wide-area service discovery and provisioning remains a challenge due to mobility of users, devices and services. Availability and failures of service instances, intermittent network disconnections, and changes in service context are just a few dynamic variations that pervasive networks have to contend with. Even if service discovery protocols are available as one means by which an application can stay informed about the current state of the environment, scalability of these pervasive systems in terms of number of services and in heterogeneity of the protocols describing and finding these services is still an obstacle. In fact, scalability, dynamicity support and enabling heterogeneity in service discovery are key objectives in wide-area service discovery. This thesis proposes a naming system to address part of these challenges in wide area service discovery. The naming system acts as a middleware mediating between services in the Clusters to enable wide-area service discovery and networking in dynamic environments. The naming system is applied to Personal Networks (PN) where user devices, nodes and services are scattered throughout personal Clusters. The services are named mainly according to their type, properties, location and ownership. The naming system resolves service names to their locations and service descriptions. The thesis starts with the design and implementation of a scalable naming system, called PNSD (P2P-based Naming System for Service Discovery), based on a structured P2P network and supporting services and network dynamics, mobility and service heterogeneity. The second part of this thesis integrates a publish/subscribe model in the naming system. The resulting system implements a notification service which informs subscribers about the services and their mobility. A security architecture is also proposed at the edge of the naming system to provide authentication and access control. As application use case, this work combines UPnP, a local-area service discovery protocol, with the PNSD system to implement a wide-area service discovery framework. Finally, the thesis focuses on the applicability of the proposed naming system to Personal Networks where a PN networking management framework is supported by the PNSD system. The latter acts as a service locating system that provides network entities with PN Cluster location information to achieve PN networking.