Multi-event algorithms and protocols for fast and robust distributed mesh provisioning and restoration
Tsuneo Nakata · Bell Labs Technical Journal · 2003
Most fast path restoration methods to date assign resources to each demand's backup path such that no contention will happen upon an expected failure event. However, it is virtually impossible for them to be prepared for multiple failures, which are to be expected in the real world, because of too many possible combinations. We have developed a novel distributed signaling algorithm that leads to best-effort restoration regardless of type and number of physical layer failures. The algorithm allows no useful resources to be left unused, while ensuring deadlock freedom. We have further generalized the algorithm to integrate parallel path search and resource reservation in case restoration attempt using preassigned backup path fails. This proves to be especially useful at an unstable state in the course of a severe failure where the resource availability changes dynamically and thus no topology information is believed to be up to date. © 2003 Lucent Technologies Inc.