Avaliando a eficácia e a eficiência da busca Par-a-Par por conteúdo

Abstract

Content-Based search on decentralized architectures, like Peer-to-Peer (P2P) Digital Libraries (DL), differs from centralized search in many aspects, including the dynamic behavior of peers and its heterogeneity in resources capacity. Evaluation of the effectiveness and efficiency of such systems is still an unsolved problem, as most of prior work either focuses only on effectiveness or considers ideal scenarios (e.g., pairs never leave the network or have unlimited bandwidth). In this dissertation, we present an evaluation of efficiency, effectiveness and the tradeoffs between these two aspects, on P2P content-based search considering several practical P2P DL aspects, such as document collection topicality, the dynamics of peer participation (churn) and peer heterogeneity (with respect to bandwidth and storage capacity). We also evaluate a document replication strategy as well as a query response merging strategy which increase the effectiveness of P2P search.