4. A case study: an Internet router

Many mission-critical applications require fault tolerance, sometimes in competition with their mission (often for performance reasons). This is particularly true of telecommunications applications, which have to meet draconian availability constraints (99.999%) that are now part of operators' basic requirements, if not simply regulatory constraints, as in the case of telephony. For this reason, fault tolerance has always been at the heart of equipment design and occupies a considerable place; it is estimated that half of a switch's software is dedicated to error and fault handling.

The field of IP (Internet Protocol) networks has not followed this path for a long time, because its very model assumes that routers are fallible and provides ways of bypassing them. As IP networks developed and became commercialized, the additional investments required by this approach...