Track:
All Papers
Downloads:
Abstract:
This paper describes a tool to diagnose the cause of failure of an HP computer server. We do this by analyzing the dump of Processor Internal Memory (PIM) and maximizing the leverage of expert learning from one hardware failure situation to another. The tool is a rule-based expert system, with some nested rules which translate to decision trees. The rules were implemented using a metalanguage which was customized for the hardware failure analysis problem domain. Pimtool has been deployed to 25 users as of December 1996. We plan to expand usage to over 400 users by end of 1997. Using Pimtool, we expect to save over 15 minutes in Mean-Time-to-Repair (MTTR) per call. We have recognized that knowledge management will be a key issue in the future and are developing tools and strategies to address it.