Técnica para obtenção de redes de autômatos estocásticos baseada em especificações de software em UML

• Ano de defesa: 2008
• Autor(a) principal: Neuwald, Felipe Barp
• Orientador(a): Fernandes, Paulo Henrique Lemelle
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Pontifícia Universidade Católica do Rio Grande do Sul; Porto Alegre
• Palavras-chave em Português: INFORMÁTICA; REDES DE AUTÔMATOS ESTOCÁSTICOS; ENGENHARIA DE SOFTWARE; SOFTWARE - AVALIAÇÃO; ORIENTAÇÃO A OBJETOS; UML (INFORMATICA)
• Link de acesso: http://hdl.handle.net/10923/1653

Performance evaluation of an object oriented software nowadays can be done, basically, by three different kinds of methods: Stress Testing, Simulations or Analytic Methods, like Stochastic Automata Networks or Queuing Networks. Stress Testing is an efficient method, but could only be performed on final phases of a software development process when the final product is almost finished. Simulations can be performed in initial phases of a software development process, but usually at a high cost, because it depends on a creation of a simulator. The Analytic Methods could be performed at the earlier phases of the software development process too, but they require a creation of a performance evaluation model. ln the object oriented software development a model is created, which contains detailed information about the system. This information could be used to generate a performance evaluation model, since many of the information containing on each other are common. Following that idea researches were made generating methods for conversion of UML diagram into performance evaluation methods, for example, to Queuing Network. However, does not exist a method or technique that converts UML to formalisms like SAN, which has a higher representative power then Queuing Networks. This work will present a technique for converting UML diagrams into Stochastic Automata Networks focusing on the representation of the logical architecture of object oriented softwares.