Aluno: Miguel Cardoso Neves
Orientador: Prof. Dr. Antonio Marinho Pilla Barcellos
Título: Enforcing properties in programmable networks
Linha de Pesquisa: Redes de Computadores
Local: Prédio 43412 – Sala 218 do Instituto de Informática
– Prof. Dr. Theophilus A. Benson (Brown University – USA – por videoconferência)
– Prof. Dr. Christian Rodolfo Esteve Rothenberg (UNICAMP – por videoconferência)
– Prof. Dr. Luciano Paschoal Gaspary (UFRGS)
Presidente da Banca: Prof. Dr. Antonio Marinho Pilla Barcellos
Abstract: Avoiding software bugs and misconfigurations in programmable networks is challenging. Recent studies show that they are among the biggest causes of failures in network infrastructures. Moreover, their consequences are normally disastrous. Previous efforts have proposed network debugging and verification techniques as a means to check that the network behaves as expected, but these techniques usually lead to incomplete solutions that can not catch all bugs or face severe scalability issues. In this thesis proposal, we introduce the abstraction of data plane monitors, special modules that allow network programmers to enforce desired properties in a scalable and expressive way. Together with P4box, a system we propose for instrumenting data plane programs with monitors, our abstraction creates an enforcement kernel that cannot be hindered, tampered or circumvented by faulty code. To assess the benefits of our mechanism, we are exploring two use cases: dynamic and static property enforcement. The former is useful when verification does not meet time constraints while the latter enables the verification of previously unfeasible properties. Preliminary results show that monitors represent a small overhead to network devices in terms of latency and resource consumption when dynamically enforcing a broad range of properties. We will focus on the static enforcement scenario during the remainder of the PhD.
Keywords: Programmable networks, network verification, network debugging, SDN, P4, monitoring.