Failure from the Field: Complexity Kills

We present a system architect and administrator's view of system dependability issues and research suggestions. The dependability problem has been extensively studied by operators and architects, and most simple failures have known solutions. However, many operators and architects are not aware of many of the available solutions or the scope of the dependability problem. More complex failures where multiple unrelated failure events cause an overall outage are becoming a more prevalent part of medium and large systems dependability in the field. Examples of these baroque multiple failure outages are presented. Implications of these two major issues on benchmarking, research, and system architecture are discussed.