Power Excursion Aware Routing in GMPLS-based WSONs F. M. V. Ramos 1 , A. Giorgetti 2 , F. Cugini 3 , P. Castoldi 2 , J. Crowcroft 1 , and I. H. White 1 1: University of Cambridge, Cambridge, UK; e-mail: fernando.ramos@cl.cam.ac.uk 2: Scuola Superiore Sant’Anna, Pisa, Italy 3: CNIT, Pisa, Italy Abstract: A routing scheme is proposed for GMPLS-based WSONs to mitigate the effects of power transients due to WDM link failures. Simulations show that power transients are considerably reduced with negligible degradation of network resource utilization. © 2008 Optical Society of America OCIS codes: (060.0060) Fiber optics and optical communications (060.4250) Networks; (060.0060) Fiber optics and optical communications, 060.4265 Networks, wavelength routing 1. Introduction Wavelength-switched optical networks (WSONs) are composed of WDM links interconnecting optical transparent nodes (e.g., optical cross connects, OXCs) able to switch the traffic directly in the optical domain. Since the WSON architecture averts the utilization of expensive opto-electronic transceivers in intermediate nodes, it is currently considered the most promising technology for next generation core and metro networks. The GMPLS control plane has been proposed for dynamic and distributed control of WSONs. In particular, GMPLS protocols provide routing, signaling and link management functions, so that in a GMPLS-based WSON end-to-end optical connections (i.e., lightpaths) can be dynamically established, maintained and released. Moreover, since in WSONs a single link failure may cause the loss of huge amounts of data, reliability represents a key network feature. The GMPLS protocols are also designed, in fact, to provide dynamic failure recovery. Many recovery schemes have been proposed in the literature, including several types of protection and restoration techniques that are currently supported by GMPLS [1]. However, all the so far proposed recovery schemes are focused on the recovery of the lightpaths directly affected by the failure, without considering the degradation that a sudden disappearance of several lightpaths (e.g., in case of a fiber cut) may introduce on surviving lightpaths. Indeed, due to the extensive usage of saturated optical amplifiers along WDM links, a sudden excursion of the passing through optical power may strongly degrade the optical signal quality of lightpaths that were partially overlapped (i.e., established along different wavelength channels along the same WDM link) with the disrupted ones. Several solutions have been proposed at the physical layer for mitigating this power excursion problem: e.g., Erbium Doped Fiber Amplifier control techniques and the utilization of link-control lasers [2]. However, these solutions considerably increase the cost and complexity of the optical amplifiers. Moreover, the aforementioned techniques have other drawbacks, such as noise figure degradation and optical gain reduction. The work in [3] tries to solve the same problem at the routing level, by introducing an integer linear programming (ILP) formulation aiming at minimizing the number of lightpaths affected by excessive power excursion in case of single link failures. However, the solution based on ILP formulation cannot scale to large networks and is not suitable in realistic dynamic scenarios (where lightpaths requests arrive one at a time) employing a distributed control plane (where each network node bases its routing decisions on the locally available network status information). This work proposes a scalable heuristic routing scheme, suitable for large GMPLS-based WSONs, aimed at mitigating the power excursion problem. We call this the Power Excursion Aware Routing (PEAR) scheme. A possible distributed implementation of the PEAR scheme is described for GMPLS-based WSONs at the protocol level, and simulations are used for evaluating the performance of the proposed scheme. 2. The Power Excursion Aware Routing (PEAR) scheme The RSVP-TE signaling protocol is used in GMPLS-based WSONs for dynamic establishment of lightpaths. The OSPF-TE routing protocol is used for distributing updated network status information among network nodes. This information is then stored in each network node in the so called Traffic Engineering Database (TED). Upon a lightpath request the source node computes a path basing its choice on information locally stored in the TED. After path computation, an RSVP-TE signaling instance is triggered along the computed path. Wavelength assignment is performed at destination exploiting the information gathered by the signaling messages. The implementation of the proposed Power Excursion Aware Routing (PEAR) scheme needs a local matrix, M, that is stored and updated in each network node by RSVP-TE and OSPF-TE protocols. Matrix M summarizes the power excursion information considering the routes of all currently established lightpaths. It is a symmetric L x L matrix, where L is the number of network links. Matrix dimensions, therefore, do not depend on the number of established lightpaths, thus preserving scheme scalability. The generic element m i,j ∈ M represents the number of wavelength channels that go down on link i in case of failure of the link j. The elements along the diagonal (i.e., m i,i ) represent the number of lightpaths established along link i. For a better understanding, Fig. 1 shows a network scenario, where