Um estudo exploratório sobre padrões de falhas de software de sistemas operacionais

• Ano de defesa: 2016
• Autor(a) principal: Santos, Caio Augusto Rodrigues dos
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Uberlândia; BR; Programa de Pós-graduação em Ciência da Computação; Ciências Exatas e da Terra; UFU
• Palavras-chave em Português: Sistemas operacionais; Falhas de software; Detecção de padrões; Estudo empírico; Causas de falha; Correlação; Sistemas operacionais (Computadores) - Avaliação; Falhas de sistemas de computação; Software - Confiança; Operating systems; Software failures; Pattern detection; Empirical study; Failure causes; Correlation; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: https://repositorio.ufu.br/handle/123456789/12592; http://doi.org/10.14393/ufu.di.2016.209

Empirical studies in software reliability have predominantly focused on end-user applications. Given the intrinsic dependency of user programs on the operating system (OS) software, OS failures can severely impact even the most reliable applications. Therefore, it is a major requirement to understand how OS failures occur in order to improve software reliability as a whole. This research carried out an exploratory study on OS failures behavior, based on 7,007 real failure records collected from different computers running a mass-market operating system. We performed a quantitative analysis to investigate different properties of the OS failures analyzed. The findings indicate that OS services failed more than any other OS failure category. We introduced an OS failure pattern detection protocol, which allow us to detect and characterize failure patterns that showed consistent across different computers from the same and varied workplaces investigated. In total, we detected 45 failure patterns with 153,511 occurrences. Based on these patterns, empirical evidences of correlation among specific OS components failures was found, which is a important property to understand better the behavior of these software failures. In addition, the proposed protocol was adapted to also detect and characterize specific causality patterns among the OS failures. Empirical evidences confirm the presence of causal correlation between the failures in the sample, where both cross-correlation and autocorrelation were found.