7.5 Protection Against Multiple Concurrent Failures

As discussed in the chapter introduction, while the network may recover from a failure very rapidly, repairing a failure may take several hours. During the repair time, additional failures may occur. Thus, demands with very high availability requirements may need protection against multiple concurrent failures. The discussion in this section focuses on recovery from two concurrent failures.

Many of the protection methods described above can, at least in theory, be extended to protection against multiple failures. For example, 1+1 dedicated protection can be enhanced to 1+2 dedicated protection, where one working path is established along with two active backup paths. If all three paths are mutually link-and-node-disjoint, then 1+2 protection can recover from any combination of two failures, where failure recovery is almost immediate. With client-side 1+2 protection, the demand source and destination both require three transponders. With network-side 1+2 protection, the source and destination both require a threeway transponder, where the transponder feeds into three different network ports (or a multicasting all-optical switch can be used). One hindrance to 1+2 protection is that in most realistic networks, there are not three totally diverse paths between all source/destination combinations. Moreover, even if three diverse paths can be found, the amount of required protection capacity would be excessive. This protection method is likely only acceptable if the number of 1+2 protected demands is very small.

(Note that the routing algorithm code provided in Appendix B is capable of finding N link-and-node disjoint paths between...