Generating recovery-command sequences by neural networks
We propose a method for automatically generating recovery-command sequences, which is intended to support quick recovery actions by system operators and to achieve automatic recovery from ICT (information and communication technology)-system failures. Our method is based on Seq2Seq (sequence-to-sequence), a neural network model usually used to solve translation tasks in the field of natural language processing. This model can learn complex relationships between logs obtained from equipment and recovery commands that operators executed in the past. When a new failure occurs, our method estimates plausible commands that recover from the failure on the basis of collected logs. Our method also evaluates the confidence score of the estimated recovery-command sequences. Operators can use this confidence score as a criterion to determine whether the estimated recovery-command sequence should be executed.
H. Ikeuchi, A. Watanabe, T. Hirao, M. Morishita, M. Nishino, Y. Matsuo, K. Watanabe, “Recovery command generation towards automatic recovery in ICT systems by Seq2Seq learning,” in Proc. of IEEE/IFIP Network Operations and Management Symposium （NOMS）, 2020, to appear.
T. Kimura, A. Watanabe, T. Toyono, K. Ishibashi, “Proactive failure detection learning generation patterns of large-scale network logs,” IEICE Transactions on Communications, Vol. E102-B, No2, pp. 306–316, 2019.