Survivability analysis on non-triconnected optical networks under dual-link failures

Survivability of optical networks is considered among the most critical problems that telecommunications operators need to solve at a reasonable cost. Survivability can be enhanced by increasing the amount of network links and its spare capacity, nevertheless this deploys more resources on the network that will be used only under failure scenarios. In other words, these spare resources do not generate any direct profit to network operators as they are reserved to route only disrupted traffic. In particular, the case of dual link failures on fiber optic cables (i.e., fiber cuts) has recently received much attention as repairing these cables typically requires much time, which then increases the probability of a second failure on another link of the network. In this context, survivability schemes can be used to recover the network from a dual link failure scenario. In this work, we analyze the case of protection and restoration schemes, which are two well-known recovery strategies. The former is simpler to implement as it considers a fixed set of backup paths for all failure scenarios; however, it cannot take into account the spare capacity released by disrupted connections. Instead, the latter computes the best recovery path considering not only the spare capacity but also the released one due to failures. Achieving 100% survivability (i.e., recovery from all possible dual link failures) requires a triconnected network, where three disjoint paths for each connection are required. Since these networks can become extremely expensive since they can require a huge number of network links (i.e., fibers connections), a more realistic case of non-triconnected networks is assumed. In these networks, full network recovery is not be feasible, but achieving the maximum possible survivability is desired. Spare capacity can then be allocated to existing network links, which represents the actual cost of the survivability. We propose optimization models that take into account these different recovery strategies, and demonstrate that restoration has the potential to provide a much better recovery capability with almost the same amount of spare capacity required in protection schemes.