Electronic Prognostics for Computer Servers

Abstract

Electronic prognostics (EP) relates measurable precursors of failures to remaining useful life (RUL) predictions for electronic systems. When used with condition-based maintenance, EP promises to significantly increase the effective mean-time-between system failures (MTBF) by predicting and avoiding system failures (maintenance planned based on condition). Increased MTBF numbers translate into enhanced availability and better operational reliability. This is essential for building highly dependable computing systems. This paper briefly reviews the steps required for achieving successful prognostics. The lessons learned from adopting EP to the previous and current generations of enterprise computing systems have influenced the next generation computer system designs, equipping future systems with more advanced capabilities for electronic prognostics. Sun Microsystems' new continuous system telemetry harness (CSTH) coupled with advanced pattern recognition substantially increases component reliability margins and system availability goals while reducing (through improved root cause analysis) costly sources of "no trouble found" events. These "no trouble found" events have become a significant warranty-cost issue for COTS and a sparing-logistics issue for mil spec electronic systems