Problem Determination, System Management and Monitoring (MELODY)
Overview
MELODY, short for IBM Mining Effectively Large Output Data Yield, is an innovative machine learning technology for analyzing large collection of log files generated by computer systems. As computers get more complex, reading their logs has become impossible for a human. Melody automatically identifies the interesting parts, discovers anomalies and allows for failure prediction, thus enable early and correct detection of malfunctions and reduce warranty costs.
MELODY-Z is a joint project with the IBM System-Z resiliency team at Poughkeepsie since 2008. MELODY-Z provide tools for anomaly detection and failure prediction in system Z logs, using machine learning and statistical analysis, and is based on MELODY-X technology. For more information see Brochure.
MELODY-X is a joint work with the IBM System-X, BladeCenter & Intellistation (xBCI) support team in Raleigh since 2005. MELODY-X is applying Machine Learning techniques on Dynamic System Analysis (DSA) dumps taken from machines under warranty and allows IBM's personnel (from front office through product field engineers (PFE) till product engineers (PE)) to view rare (and therefore probably interesting) log events, unique (and therefore probably important) configuration items as well as output of rules designed to help pin-point the problems better and faster. We also explore ways to propose data driven rules to the engineers. For more details, see the following summary.
Contact: Ofer Lavi (oferl@il.ibm.com)