TMACS: Type-based distributed middleware for mobile ad-hoc networks

Mobile Ad-hoc Networks (MANETs) are an emerging wireless networks platform for deploying diverse mobile applications ranging from battlefield, disaster relief to ad-hoc conferencing. This dissertation presents the design and implementation of TMACS---a middleware framework to simplify the development and deployment of group-oriented ad-hoc services and applications in MANETs.;TMACS is built upon the novel concept of type-based group communication paradigm in which type is used as a first-class abstraction for identifying groups; object-oriented principles such as subtyping and type-composition are used to establish hierarchical group relationship and dynamically control communication scope. TMACS provides a distributed computing model called TRPC to let applications invoke a remote method call on a group of distributed objects simultaneously. The group is identified based on the combination of type and scope---an abstraction for non-type properties. The method calls of TRPC are non-blocking and the caller can retrieve the return values asynchronously. A fully decentralized discovery service has been developed in TMACS to lookup the meta-information of the distributed objects and services. To efficiently support TRPC and service discovery at network layer, we propose and design TypeCast routing protocol to perform effective type dissemination and aggregation in a self-organized routing topology that is resilient to mobility and link failure.;As case studies, we implement an ad-hoc distributed caching service and an ad-hoc marketplace application with TMACS. The caching service enables mobile nodes to cache arbitrary service objects among each other to increase service availability and fault tolerance; the ad-hoc marketplace provides an auction platform for peers in MANETs to trade items in real-time without relying on centralized brokers. We conducted detailed simulation study and physical experiments to evaluate the performance of TMACS and demonstrate its scalability and resiliency in the presence of mobility-induced topology changes.