Diverse Routing Strategies for On-Demand Lightpath Provisioning in ASON/GMPLS Networks Luis Velasco, Salvatore Spadaro, Jaume Comellas, Gabriel Junyent Optical Communications Group - Universitat Politècnica de Cataluya (UPC), C/ Jordi Girona, 31. 08034 Barcelona Spain E-mail: {luis.velasco, spadaro, comellas, junyent}@tsc.upc.edu The average increment and the increasing variability of the traffic to be carried by ASON/GMPLS transport networks will require on-demand provisioning in the near future. In this scenario dynamic routing based on traffic engineering techniques is a key issue. This paper shows through simulation experiments how using network failures information with dynamic routing in ASON/GMPLS networks can considerably impact on the performance of such network. Three routing strategies are considered differing only in its use of failure information. The performance is evaluated in terms of request blocking probability and availability over two different networks from a sparse one to another highly meshed. This work has been partially founded by RINGING project (TEC2005-08051-C03-02) from MEC (Spanish Government). 1. INTRODUCTION The introduction of intelligence in Automatically Switched Optical Networks (ASON) [1] using a GMPLS control plane [2] allows provisioning of optical channels (lightpath) in seconds. This would open new opportunities related to better resource utilization, creation of new services, such as bandwidth on demand, and a range of traffic engineering mechanisms [3]. Client network layers (IP, SDH, etc) can request optical connections to the optical network through the User Network Interface (UNI) [4]. On the other hand, optical transport networks provide a huge amount of bandwidth in each lightpath connecting two main locations. In order to maximize the occupation of the lightpaths it is necessary to perform grooming. This groups together different client layer demands between locations. The difference of scale between lightpath and typical client layer demand bandwidth make lightpaths to be very stable in the time and are mainly permanent connections. Under these considerations, it is clear that the only way to improve availability in optical transport networks is by the means of recovery (protection or restoration) schemes. So, a lot of effort has been made to create, compare, develop, etc. different recovery schemes [5]. But the scenario is rapidly changing, the traffic to be carried by today’s backbone networks increases very rapidly mainly due to the massive use of internet and multimedia applications. The difference of scale between transport and client bandwidth is falling down and the need to establish lightpath in an on-demand way as switched connections is rising. On the other hand, there are applications that need a huge amount of bandwidth between two locations for a determinate period of time. Examples of these applications are HDTV [6], Grid computing [7], Tele-immersion [8], etc. Switched connections are more sensible about failures in the network than permanent connections because they are on-demand requested when they are really needed and much of the times would require the highest level of availability. Therefore, it is necessary to periodically update routing tables to avoid congestion and failures for new lightpath requests, allowing dynamic routing based on traffic engineering techniques, to find the best route in the network. In an ASON/GMPLS based network, this information can be advertised in OSPF TE LSAs [9]. 141