A toolkit for event analysis and logging

This report describes the Toolkit for Event Analysis and Logging that came out of IBM's effort to converge its support for low level HPC event analysis. The toolkit is designed as a pluggable processing pipeline that allows different components to use connectors to log events, have analyzers evaluate the events to get closer to the root cause, report and log these as alerts, and deliver the alerts to interested parties via filters and listeners. A number of plug-ins are provided with the toolkit, including a filter that removes duplicate alerts and a rule driven event analyzer, which has a new rules language that allows for a higher level location centric specification of event analysis. In order to work with the HPC community to enhance the toolkit and expand it to more platforms, it is available as open source on sourceforge.