We consider non-degradable hierarchical redundant systems having multiple working and failure modes with restoration time depending on failure type. We evaluate these systems using two measures: generalized uptime and traditional downtime. We define the Impact Weighted System Uptime (IWSU) and illustrate its usefulness in practical terms, viz., an IP router. Next, we provide an analysis that fits the downtimes by a heavy-tailed log PH distribution. For these downtime distributions, we study

more »

... her it is more cost effective to reduce failure rates or to speed up the response to failures The first option is a vendor problem, but the second is a service provider problem. A numerical example is given to help appreciate the tradeoff.