For the communications among different carriers in the backbone optical networks, IETF has proposed the routing and signaling protocol, Optical Border Gateway Protocol (OBGP). However, the fast protection and restoration scheme in mesh networks by using OBGP has not been studied. One way adopts the local Routing Information Base to obtain a pair of lightpaths between the source and destination Autonomous Systems (AS's). The lightpath with an Optimal Path Identifier acts as the working path and the other one is the dedicated backup path. When the working path fails, the lightpath establishment along the backup path applies the Serial Restoration Scheme (SRS), in which the message for resource reservation is forwarded to the upstream AS along after the Optical Cross Connects (OXC) is reconfigured.We propose a fast protection switching scheme, Parallel Restoration Scheme (PRS), in which the reservation message is transmitted ahead of OXC reconfiguration. For the single-link failure, the total protection switching time includes the time of detecting the failure in the upstream AS of the failed link (T1), the time of notifying the Source AS (S-AS) about the failure (T2), and the time of requesting and reservation to setup the backup path between S-AS and Destination AS (D-AS) (T3 and T4). The SRS and PRS have the same T1, T2, and T3, and the latter two are determined by the signaling processing delay per node (T5), the signaling transfer delay per link (T6), and the length of the notifying span along the working path or the length of the backup path (P). In the SRS, when the reservation message arrives, each AS along the backup path performs the OXC reconfiguration with the switching time (T7) and then forwards the message to the upstream AS, so T4 is formulated as (P-1)*(T5+T6)+P*T7. In the PRS, when receiving a reservation message, each AS transmits it to the upstream AS immediately, and then makes local hardware reconfiguration, so T4 is expressed as (P-1)*(T5+T6)+T7. The procedure of PRS can be seen in the Figure.We made the simulations on four mesh topologies with the number of AS's 5, 13, 25, 41, in which the failure occurs in a random link along the working path for 1000 pairs of working and backup paths. We assume T1, T5, T6 and T7 are 5ms, 0.2ms, 2ms and 20ms, respectively, and conclude that the mean protection switching time in the PRS is much shorter than that in the SRS, and more obvious advantage can be gained in larger network scale. The ratio of PRS to SRS is less than 50% for all the topologies and reaches less than 40% for the last one. The PRS with the mean protection switching time of 40 to 50ms meets the demand of less than 50ms.
© 2006 Optical Society of America