Arquitetura ciente de contexto para aplicações sociais móveis

• Ano de defesa: 2012
• Autor(a) principal: Rafael Guimaraes Siqueira
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: Computação ubíqua; Redes sociais móveis; Middleware; Arquitetura de redes de computador; Computação; Sistemas de comunicação
• Link de acesso: http://hdl.handle.net/1843/ESBF-8UZH6X

Mobile social applications explore the potential of social networks on mobile platforms. They represent peer-to-peer applications, for their social aspect and the individuality of mobile devices. Nonetheless, in many cases, they are still deployed using a client-server model. In practice, this means that communications between two devices physically collocated happens through servers located in other continents, resulting in higher response times. On isolated cases in which these applications are deployed using peer-to-peer models, they use communication protocols inappropriate for environments where node content is constantly changing, such as user contexts on mobile computing environments. To implement social applications for these mobile devices, amongst them smartphones and tablets, it is ideal that their underlying infrastructure is created, from its conception, to support communications on mobile social networks. This work presents an architecture that provides such infrastructure. It proposes the creation of a context-aware peer-to-peer overlay that clusterizes users, whose probability of interaction on the social network is higher. Factors which, potentially, increment this probability are the pre-existence of an explicitly defined relationship or the coincidence of characteristics which could catalyze the establishment of this relationship in the future. The proposed protocol in this architecture uses descriptions of knowledge and user contexts as logic predicates, from which it is possible to infer other contexts more useful from the applications point of view. This approach is particularly flexible because it allows for the representation of relevant knowledge from many different problems, by means of these logic predicates. Algorithm scalability is tested with experiments on a discrete event simulator, which allows for empirical analysis of impact on individual nodes. Furthermore, the viability of the solution is demonstrated by an example mobile social application for campus community living, tested using a dataset with mobility traces of campus inhabitants. The architecture implemented in this work contributes for the development of mobile social applications more adequate to the ubiquitous computing paradigm, by providing a self-organizing infrastructure more suited for these applications and a context-aware protocol that organizes them in that manner.