Detalhes bibliográficos
Ano de defesa: |
2014 |
Autor(a) principal: |
ANTONINO, Pedro Ribeiro Gonçalves |
Orientador(a): |
Não Informado pela instituição |
Banca de defesa: |
Não Informado pela instituição |
Tipo de documento: |
Dissertação
|
Tipo de acesso: |
Acesso aberto |
Idioma: |
eng |
Instituição de defesa: |
Universidade Federal de Pernambuco
|
Programa de Pós-Graduação: |
Não Informado pela instituição
|
Departamento: |
Não Informado pela instituição
|
País: |
Não Informado pela instituição
|
Palavras-chave em Português: |
|
Link de acesso: |
https://repositorio.ufpe.br/handle/123456789/11966
|
Resumo: |
The increase of computer systems complexity has led to a direct increase in the difficulty of verifying their correctness. For mastering this complexity, formal methods can be used in the development of systems providing techniques for both design and verification. Regarding concurrent and distributed systems, the necessity of a formal approach is more prominent given the substantial increase in complexity due to the countless number of interactions between their constituent systems. Unfortunately, however, current methods are not capable of dealing with the automated analysis of such systems in general, even if we consider only classical properties such as deadlock freedom; the state explosion problem is the main reason for this ineffectiveness. This work is a contribution in this direction. Particularly, considering networks of CSP processes, this work proposes a local strategy for deadlock analysis based on the notion of process refinement. The locality of this strategy prevents the state explosion problem generated by the interaction of constituent systems, which represents a major asset of our strategy. We define a refinement assertion for checking conflict freedom between pairs of processes in the network; this can be used for the local verification of networks with an acyclic communication topology. Concerning networks with a cyclic communication topology, we propose three patterns that prevent deadlocks: the resource allocation, the client/server and the async dynamic. These patterns impose behavioural and structural restrictions to prevent deadlocks. The behavioural restrictions are also captured by refinement assertions, which enable one to automatically verify these conditions using a refinement checker. Besides this, we develop four case studies to evaluate the efficiency of our strategy in practice: a ring buffer, a dining philosopher, and two variations of a leadership election algorithm. One of the variations of the leadership election algorithm consists of a model used in practice by the B&O Company, an industrial partner. In this study, we compare our strategy with two other techniques for deadlock freedom verification, the SSD algorithm of the Deadlock Checker tool and the built-in deadlock freedom assertion of FDR. This study demonstrates how our strategy can be used and that it might be a useful alternative to analysing complex industrial systems for deadlock freedom. |