Automatic Anomaly and Failure Detection
The pervasive population growth of CPUs and executing programs has outpaced our ability to manage digital systems without automated assistance. Ideally, digital systems will become aware of their own behaviors and contribute to their own maintenance. This project takes a step in this direction by providing a framework for building statictical models of aggregate program behavior from data collected during the executions of an instrumented program. The project will builds and evaluates behavior classifiers from these data to use in predicting the behavior of new program executions.