Logs are a valuable source of clues about why a system has misbehaved or failed. There have been past studies about using logs to automatically troubleshoot customer problems and these studies have shown that this is very challenging for several reasons. First, log messages tend to be an afterthought and are often poorly formatted. Second, it is difficult to understand log messages across modules. Third, log messages might miss vital information that could have helped resolve a problem quickly.
This project will study how to enhance log messages so that they aid in the quick and automatic resolution of customer problems.