Designing for Ultrahigh Availability: The Unix RTR Operating System

Early designers of highly available computers concentrated on recovery from hardware failures to keep downtime to a minimum. But as hardware became more reliable, and systems (particularly software) became more complex, the percentage of downtime caused by hardware decreased. Achieving ultrahigh availability-on the order of a few minutes of downtime per year-requires far more than just reliable hardware. This can be seen from Table 1, which gives causes of downtime for both electronic switching 1-3 and transaction processing.4 Although the numbers differ, in both cases the hardware accounts for less than half the downtime. The other causes include the following: