Um modelo multiperspectiva para avaliação de desempenho de plataformas de processamento de grafos

• Autor(a) principal: Silva, Daniel Nascimento Ramos da
• Data de Publicação: 2017
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações do LNCC
• Texto Completo: https://tede.lncc.br/handle/tede/254
• topic: Teoria dos grafos; Processamento de grafos; Theory of graphs; Graph processing; CNPQ::ENGENHARIAS::ENGENHARIA DE PRODUCAO::PESQUISA OPERACIONAL::TEORIA DOS GRAFOS

Many emerging challenges currently arising in science relate to understanding the dynamics and structure of complex systems consisting of interacting components. People on online social networks, power grids, air transportation networks, human brain connections, and agents in the financial market are some examples of network systems present in many domains. These systems, given their scale and non-trivial connectivity patterns, are called complex networks. In this sense, the dimension of these networks turns imperative the adoption of computing systems in support to their analysis. In this case, graphs are the usual tool for the representation of complex networks. Moreover, as the scientific investigation for these graphs is of great significance for many domains, the researcher and developer communities have proposed many computing platforms for their processing. However, the multitude of graph processing platforms brings up questions about the implications of their adoption, given the analysis, network and computing environment characteristics of the analysts. Therefore, in this dissertation, we propose a multi-perspective model for the performance evaluation of graph processing platforms which differs from related works in simultaneously approaching the topic from four perspectives: algorithms, computing environment, networks, and platforms. Additionally, we carry out a performance evaluation study that follows the directives indicated by the proposed model for a representative set of algorithms, platforms, and networks, demonstrating the proposed model applicability as exhibits the relationship between algorithms, networks and platforms efficiency.