Once failures occur in a cloud datacenter accommodating a large number of virtual resources, they tend to spread rapidly and widely, impacting many cloud services and their users.One of the best ways to prevent a failure from spreading in the system is to identify signs of a failure before its occurrence and deal with it proactively before it causes serious problems.Although several approaches have been proposed to predict failures by analyzing past logs of system messages and identifying the relationship between the messages and the failures, it is still difficult to automatically predict the failure for several reasons such as variation of log message formats and frequent changes in their configurations.Based on this understanding, we propose a new failure prediction method that Fujitsu Laboratories has developed.The method automatically learns message patterns as signs of failure by classifying messages by their similarity regardless of their format and re-learning the message patterns in frequently changed configurations.We evaluated our method in an actual cloud datacenter.The experimental results showed that our approach predicted failures with 80 precision and 90 recall in the best case.
展开▼
