System and method for trend estimation for application-agnostic statistical fault detection

摘要

Trend estimation for application-agnostic statistical fault detection of multi-process applications in environments with data trend includes at least one of: a multi-process application runs on a host. Statistical events are collected and sent to a statistical fault detector. The statistical fault detector creates one or more distributions and compares recent statistical event data to historical statistical event data and uses deviation from historical norm for fault detection. Trend is estimated, and if needed, removed from event data prior to the creation of distributions. Trend is estimated using spectral techniques, filter banks and Maximum Entry Spectral Estimation, and dominant frequencies are estimated and utilized to adapt to the environment.