Single-layer versus multilayer preplanned lightpath restoration

Abstract

Upon failure occurrence, preplanned lightpath restoration recovers a faulted lightpath by selecting the backup path from a set of preplanned paths. Preplanned lightpath restoration combines the fast recovery time guaranteed by off-line backup path computation with the adaptive recovery guaranteed by the on-line choice of the backup path. One of the main issues in preplanned lightpath restoration is the backup path selection. To address this problem, stochastic preplanned restoration schemes have been proposed. In this study, a family of deterministic preplanned restoration schemes is introduced and evaluated. The deterministic preplanned restoration with proportional weighted path choice (DPR-PW) and the multilayer DPR-PW (MDPR-PW) schemes are proposed as single-layer (e.g., optical layer) and multilayer (e.g., sublightpath granularity) preplanned restoration schemes, respectively. Both schemes utilize the same network state information employed in stochastic preplanned restoration schemes. The originality of both DPR-PW and MDPR-PW is the deterministic choice of the paths for recovering disrupted lightpaths. The choice minimizes the error between the fraction of lightpaths assigned to each backup path and the backup path selection probabilities that, in turn, are computed to minimize resource contention, as in the stochastic schemes. Numerical results show that, if lightpath capacity is fully utilized, then DPR-PW, MDPR-PW, and multilayer stochastic preplanned restoration schemes achieve similar performance in terms of fraction of recovered connections. However, while both deterministic and stochastic multilayer schemes do require network nodes to implement sublightpath connection grooming capabilities, the DPR-PW scheme does not. Thus, DPR-PW optimizes recovery performance by slightly increasing the intranode coordination required during lightpath recovery with respect to stochastic schemes but without impacting the overall network cost. Finally, in all the considered schemes, the aggregation of the signaling necessary for reserving and activating spare bandwidth along backup paths allows to limit the signaling overhead while negligibly increasing the recovery time.