Heuristic Solution to Protect Communications in WDM Networks using P-cycles

Optical WDM mesh networks are able to transport huge amount of information. The use of such technology however poses the problem of protection against failures such as fibre cuts. One of the principal methods for link protection used in optical WDM networks is pre-configured protection cycle (p-cycle). The major problem of this method of protection resides in finding the optimal set of p-cycles which protect the network for a given distribution of working capacity. Existing heuristics generate a large set of p-cycle candidates which are entirely independent of the network state, and from then the good sub-set of p-cycles which will protect the network is selected. In this paper, we propose a new algorithm of generation of p-cycles based on the incremental aggregation of the shortest cycles. Our generation of p-cycles depends on the state of the network. This enables us to choose an efficient set of p-cycles which will protect the network. The set of p-cycles that we generate is the final set which will protect the network, in other words our heuristic does not go through the additional step of p-cycle selection

💡 Research Summary

The paper addresses the problem of fast and efficient protection in wavelength‑division‑multiplexed (WDM) mesh optical networks using pre‑configured protection cycles (p‑cycles). A p‑cycle combines the advantages of ring protection with the ability to protect not only the links that form the cycle (on‑cycle links) but also any “straddling” links whose end nodes lie on the cycle. The main challenge is to select an optimal set of p‑cycles that can protect a given working capacity distribution; this problem is NP‑hard and the number of possible p‑cycles grows exponentially with network size.

Existing heuristic approaches typically consist of two stages: (i) generation of a large pool of candidate p‑cycles based solely on network topology, independent of the current traffic load, and (ii) selection of a subset of these candidates using integer linear programming (ILP) or a greedy heuristic. Because the candidate generation ignores the actual working capacity on each link, many generated cycles are inefficient, and the final selected set can be large, leading to high configuration complexity and sub‑optimal resource utilization (high redundancy).

The authors propose a novel heuristic that integrates candidate generation and selection into a single process driven by the network state. The method proceeds in two steps. First, all “shortest cycles” are generated: for each link X, the shortest path between its end nodes that does not use X is found, and X is added to this path to form a simple cycle without straddling links. Redundant cycles that cannot be improved are immediately placed in the final p‑cycle set and removed from the graph.

Second, the algorithm builds high‑efficiency p‑cycles by incrementally aggregating these shortest cycles. It repeatedly selects a cycle (Cyc) that contains a link with the smallest positive working capacity (w_j). Then it searches for another shortest cycle (Cyc_i) that satisfies three strict conditions: (1) Cyc and Cyc_i share exactly one link, (2) they share no nodes other than the two end‑nodes of the shared link, and (3) the redundancy (total spare capacity divided by total working capacity) after merging is lower than before. If such a Cyc_i is found, the two cycles are merged, creating a new cycle with one additional straddling link and a higher protection capability (n_i wavelengths on the shared link). The new merged cycle becomes the current Cyc, and the process repeats until no further merge satisfies the conditions. After a high‑efficiency p‑cycle is constructed, the working capacity it protects (n_i on‑cycle links and 2 × n_i on straddling links) is removed from the network, and the algorithm restarts to protect the remaining traffic.

The authors evaluate the approach on a 28‑node, 45‑link US long‑haul network, with traffic routed on shortest paths and full wavelength conversion assumed at each node. Compared with the Capacitated Iterative Design Algorithm (CIDA), the proposed heuristic consistently achieves lower redundancy—always below 100 %—and reduces the number of p‑cycles in the final protection set by more than 30 %. The reduction stems from the ability of the generated p‑cycles to detour multiple wavelengths simultaneously, whereas CIDA’s cycles detour only one wavelength at a time.

Key contributions of the paper are: (1) a state‑aware generation of p‑cycles that directly incorporates working capacity information, (2) elimination of the separate candidate‑selection phase, resulting in a simpler and faster design process, and (3) empirical evidence of superior resource utilization and configuration simplicity compared with a well‑known heuristic.

The paper also acknowledges limitations: the iterative shortest‑cycle search and merging steps involve graph traversals that may become costly for very large topologies. Future work is suggested to explore parallel or approximate shortest‑path techniques to accelerate the algorithm, and to extend the framework to multi‑level protection schemes that consider quality‑of‑service constraints. Overall, the work presents a practical and effective heuristic for p‑cycle‑based protection that adapts to the actual traffic state of WDM networks.