Contributions to presence-based systems for deploying ubiquitous communication services

Resumen

Next-Generation Networks (NGNs) will converge the existing fixed and wireless networks. These networks rely on the IMS (IP Multimedia Subsystem), introduced by the 3GPP. The presence service came into being in instant messaging applications. A user¿s presence information consists in any context that is necessary for applications to handle and adapt the user's communications. The presence service is crucial in the IMS to deploy ubiquitous services. SIMPLE is the standard protocol for handling presence and instant messages. This protocol disseminates users' presence information through subscriptions, notifications and publications. SIMPLE generates much signaling traffic for constantly disseminating presence information and maintaining subscriptions, which may overload network servers. This issue is even more harmful to the IMS due to its centralized servers. A key factor in the success of NGNs is to provide users with always-on services that are seamlessly part of their daily life. Personalizing these services according to the users' needs is necessary for the success of these services. To this end, presence information is considered as a crucial tool for user-based personalization. This thesis can be briefly summarized through the following contributions: We propose filtering and controlling the rate of presence publications so as to reduce the information sent over access links. We probabilistically model presence information through Markov chains, and analyzed the efficiency of controlling the rate of publications that are modeled by a particular Markov chain. The reported results show that this technique certainly reduces presence overload. We mathematically study the amount of presence traffic exchanged between domains, and analyze the efficiency of several strategies for reducing this traffic. We propose an strategy, which we call Common Subscribe (CS), for reducing the presence traffic exchanged between federated domains. We compare this strategy traffic with that generated by other optimizations. The reported results show that CS is the most efficient at reducing presence traffic. We analyze the load in the number of messages that several inter-domain traffic optimizations cause to the IMS centralized servers. Our proposed strategy, CS, combined with an RLS (i.e., a SIMPLE optimization) is the only optimization that reduces the IMS load; the others increase this load. We estimate the efficiency of the RLS, thereby concluding that the RLS is not efficient under certain circumstances, and hence this optimization is discouraged. We propose a queuing system for optimizing presence traffic on both the network core and access link, which is capable to adapt the publication and notification rate based on some quality conditions (e.g, maximum delay). We probabilistically model this system, and validate it in different scenarios. We propose, and implement a prototype of, a fully-distributed platform for handling user presence information. This approach allows integrating Internet Services, such as HTTP or VoIP, and optimizing these services in an easy, user-personalized way. We have developed SECE (Sense Everything, Control Everything), a platform for users to create rules that handle their communications and Internet Services proactively. SECE interacts with multiple third-party services for obtaining as much user context as possible. We have developed a natural-English-like formal language for SECE rules. We have enhanced SECE for discovering web services automatically through the Web Ontology Language (OWL). SECE allows composing web services automatically based on real-world events, which is a significant contribution to the Semantic Web. The research presented in this thesis has been published through 3 book chapters, 4 international journals (3 of them are indexed in JCR), 10 international conference papers, 1 demonstration at an international conference, and 1 national conference