Designated PCE Election Procedure for Traffic Engineering Database Creation in GMPLS Multi-Layer Networks F. Cugini (1) , N. Andriolli (2) , G. Bottari (3) , P. Iovanna (3) , L. Valcarenghi (2) , and P. Castoldi (2) (1) CNIT, Pisa, Italy, filippo.cugini@cnit.it (2) Scuola Superiore Sant’Anna, Pisa, Italy (3) Ericsson, Pisa, Italy Abstract A Designated PCE (DP) election procedure for PCE Traffic Engineering Database creation and maintenance is proposed. A DP-based scheme is applied to improve the network stability and scalability while guaranteeing effective multi-layer Traffic Engineering performance. Introduction Path Computation Element (PCE) Architecture has been defined to compute and provide effective Traffic Engineering solutions. To this end, an accurate and timely PCE TE Database (TED) is required [1]. Traditionally, the PCE TED has been retrieved from a link state routing protocol (e.g., OSPF-TE). However, the amount of TE information to account for may be extremely high, particularly in the case of detailed WSON information [1] or, as considered in this study, of Forwarding Adjacency Label Switch Paths (FA-LSPs) information in GMPLS multi-layer networks (MLN) [2]. The advertisement of this kind of information through a routing protocol may determine convergence and scalability issues and may affect the processing, storage and communication performance of network nodes. In [1], three alternative methods to create and maintain a PCE TED are investigated: (1) nodes send local information to all PCEs; (2) nodes send local information to an intermediate server that will relay it to all PCEs; (3) nodes send local information to at least one PCE and have the PCEs share this information with each other. In [1], due to the informational nature of the document, no implementation details are provided. No practical solutions have been proposed so far, especially in the context of GMPLS MLN. In this study, we focus on the implementation of a slightly upgraded version of the third method, where TE information is sent to at least two PCEs for reliability purposes. 2. Forwarding Adjacency (FA) in PCE-based GMPLS Multi-layer Networks The considered GMPLS multi-layer network (MLN) scenario, in which the proposed method is applied, consists in a two-layer network, where layers are characterized by the same Interface Switching Capability (ISC) and are referred to as lower (e.g., ISC of type Lambda Switching Capable (LSC)) and upper (e.g., ISC of type Packet Switching Capable (PSC)). In compliance with the GMPLS MLN specifications [2]: (i) a single GMPLS control plane instance is considered; (ii) a Label Switched Path (LSP) starts and ends at the same layer (i.e., ISC); (iii) once an LSP is established at the lower layer from one layer border node to another, it can be used as a data link in the upper layer. Furthermore, an LSP at the lower layer can be advertised as a TE Link and exploited in the path computation of LSPs originated by different nodes. Such TE Link is referred to as FA-LSP. An FA-LSP has the special characteristic that it does not require the set up of a routing adjacency (peering) between its end points. At the upper layer, the FA-LSPs compose the Virtual Network Topology (VNT) provided by the lower layer. The VNT facilitates the path computation of LSPs in MLN since it describes the resources at a single layer. In this way, per- layer path computations can be rapidly performed without considering the whole MLN TED. To address scalability and reliability requirements, multiple PCEs per layer are typically considered: L-PCE i and U-PCE j are responsible for path computations at the lower and upper layer respectively (2≤i≤M, 2≤i≤N). The PCE TEDs are retrieved from the routing protocol, e.g. by listening to the OSPF-TE advertisement. It is an implementation decision whether or not the whole VNT is advertised and made available in the path computation of LSPs originated by different nodes. Two main implementation schemes have been discussed considered so far. In the first scheme, here referred to as No-FA, the FA-LSPs are not advertised at the upper layer. In No-FA, upon connection request from source s to destination d, U-PCE j performs the path computation by exploiting just FA-LSPs starting at node s and terminating at node d. If this computation fails because of lack of resources, U-PCE j requests L-PCE j to compute a new segment or path at the lower layer, which is then exploited to ECOC 2010, 19-23 September, 2010, Torino, Italy 978-1-4244-8535-2/10/$26.00 ©2010 IEEE Tu.4.B.5