Dynamic Reconfiguration and Survivability 
in WDM Optical Networks

Wavelength-division multiplexing (WDM) and wavelength routing are becoming the technology-of-choice for use in the next generation backbone transport networks. A WDM optical network can utilize the large bandwidth available in an optical fibre to realise several hundreds of channels, each at a different optical wavelength. In a WDM optical network, optical cross-connects (OXCs) are interconnected by point-to-point fibre links. Here, a lightpath is used for transmitting a message optically between its end nodes. 

In a WDM-based transport network, the WDM optical layer provides lightpath services to the client layer such as Internet Protocol (IP) / multi-protocol label switching (MPLS) (Figure 1). The set of lightpaths in the optical layer defines the virtual topology. When the traffic demand pattern changes in the IP layer, the network performance may become poor. In order to improve the network performance, the virtual topology can be reconfigured to suit the changing traffic patterns. Such a reconfiguration is feasible as the optical cross-connects are reconfigurable.

This work uses virtual topology reconfiguration in WDM optical ring networks. The ring is an attractive architecture for use in regional and metropolitan area networks (MANs) covering a small geographical area. A reconfiguration algorithm called Merge Split Reconfiguration (MSR) that restricts the virtual topology evolution to reduce the network disruption and reconfiguration cost has been developed. The algorithm is based on the concept of splitting and merging of existing lightpaths. To set up new lightpaths between node pairs, some of the existing lightpaths can be merged or split to efficiently utilize resources such as wavelengths, optical transmitters and optical receivers. When we attempt to reduce the reconfiguration changes, lightpath load (and hence congestion) in the network may increase. Reconfiguration using the MSR algorithm results in low network congestion and reduced number of lightpath changes. Further, the MSR algorithm is computationally simple. It ensures that the number of optical cross-connects that need to be configured is limited. 

Figure 1: IP/MPLS-over-WDM networks.


Figure 2: Blocking probability versus offered load for different schemes.

In WDM networks, survivability becomes an important issue since a single fibre cut can cause a large amount of traffic loss. When a new label-switched path (LSP) request arrives with a specific bandwidth requirement, a primary LSP and a link-disjoint backup LSP need to be chosen. We propose two integrated routing algorithms: hop-based integrated routing algorithm (HIRA) and bandwidth-based integrated routing algorithm (BIRA) to select the primary and backup LSPs. The algorithms basically model the network as a graph, assign weights to different edges, and use a shortest-path selection algorithm such as Dijkstra's algorithm to choose the primary and backup LSPs. Based on the cost metric, such as the number of hops and amount of bandwidth, they determine whether to route a connection request on the existing lightpaths, open new lightpaths, or use some existing lightpaths and create additional ones. 

To optimize the network resources, we consider integrated-routing of restorable connections with backup sharing capabilities. We compare our approach using integrated routing to two baseline schemes: WDM shared protection and sequential routing protection approach. In the WDM shared protection scheme, each lightpath is protected by a backup lightpath and backup lightpaths are allowed to share wavelengths whenever possible. In the sequential routing approach, the primary and backup LSPs are first routed in the IP/MPLS layer and only if it fails, a new lightpath at the optical layer is created. As shown in Figure 2, the two proposed approaches outperform the existing schemes considerably in terms of connection blocking probability.