A Novel Optical Layer-Based Restoration Approach for IP-Over-WDM Networks

The optical layer can restore the disrupted tralfic on a persall basis in the case of IinWrouter failures. Thus, the optical layer can provide different levels of restoration for different classes of service. IINTRODUCTION 8. MOTIVATION There is an emerging consensus that IP-over-WDM networking architecture is the most suited for the transport and protection of the rapidly growing IP traffic. In this scenario, the role of SONET/SDH will diminish, and 1P networks will evolve towards a model that consists of high-performance IP/MPLS routers attached to an optical transport network. Several architectural options have been proposed on how the IP routers must interact with the optical layer to achieve end-to-end connectivity, including overlay, augmented, and peer-to-peer models [I-21. One of the most essential elements of 1P-over-WDM backbone network design is to provide resilience to failures such as fiber-optic cable breaks and router failures. Restoration may be provided at the IP layer andor the optical layer. With IP layer restoration in response to a router-router link (logical link) or node (router) failure, IP routers reroute packets onto new IP logical links. With optical layer restoration, a significant number of failure scenarios including fiber cable breaks and transmission equipment outages can be detected and recovered much quicker than IP rerouting. These failures can be restored transparently (core routers are not notified of the fault) and therefore do not cause 1P routing reconvergence. Furthermore, link failure restoration at the optical layer is expected to be more cost effective than either the SONETISDH layer or the IP layer [3-51. Although the above discussion may suggest that optical layer restoration is certainly a desirable feature of IP backbone circuits, in practice this is not currently the case particularly when router failures are taken into consideration. Protection against router failures (that constitute the majority of faults in many backbone networks [3]) must be provided directly at the IP layer, and IP network operators have to provide extra link capacity to recover from such failures. Furthermore, different classes of emerging IP services need varying degree of resilience requirements. However, only two protectionirestoration classes are possible at the optical layer, unprotected or protected full wavelength traffic. These are the main motivations that have always compelled IP network operators to exclusively adapt an IP layer-based restoration strategy when router failures are taken into account. If the optical layer was capable of coping with router failures, an independent optical layerbased restoration strategy would be much more beneficial for both 1P and transport network managers. This paper proposes a novel optical layer-based resilience approach for the emerging IP-over-WDM networking model. In this approach, the optical layer can restore the disrupted traffic on a per-call basis in the case of link and router failures. 11OPTICAL LAYER-BASED UNIFIED CONTROL PLANE ARCHITECTURE To implement the proposed networking resilience approach, this work proposes a novel IP-over-optical network interconnection model utilizing an optical layerbased unified control plane that manages both routers and optical switches. In this architecture all of the networking intelligence (both physical and logical connectivity) is shifted to the IPMPLS-aware, non-traffic bearing OXC controller modules. This model enables the optical layer to independently provision a full range of bandwidth entities, e.g. low-rate traffic streams (sub-lambdas), and full wavelengths and to provide selective restoration and QoS granularity. The OXC controller is also responsible for creating, maintaining and updating both the physical and logical connectivity tables. The responsibility of the edge IPiMPLS router is then simply to request a service from the OTN and the latter is responsible for providing this service. GMPLS can support the unified control plane to provide Iambdahb-lambda routing, signaling and survivability functionalities in the optical domain. 111OPTICAL AYER-BASED RESTORATION Restoration of low-rate traffic streams can now be accomplished at the optical layer and the burden of maintaining a database for keeping track of these “sublambda” connection requests can also be shifted to the optical layer. The feasibility of an optical layer capable of provisioninghestoring on a per-call basis introduces the capability of selective restoration, which in turn allows for different levels of restoration (differentiated resilience) for different classes of service. A. Edge Router Failure In the proposed approach the affected traffic traversing the failed router can be restored by rerouting the individual affected calls. In addition, all lightpaths originating or 0-7803-7888-1/03/$17.~03 IEEE 959 terminating at the failed router can be immediately released so that the resources can be made available for future connections. Fig. 1: Edge Router Failure As an illustrative example (Fig. I), if router E fails, the lightpath from A to E and the lightpath from E to F are released. If there were calls utilizing these two lightpaths sequentially (i.e. traffic from A to F, through E), they can be restored on a per-call basis. For example, these calls can be serviced by setting-up a new lightpath directly from A to F (lightpath on the path A-C-E-F, not shown). B.~Physical Link Failure(Trunk cut) In the case of a trunk cut,.all affected calls appear to the network as new calls;-information about how these calls were previously serviced is erased and only the call characteristics are kept (i.e. the source, the destination, the bandwidth requirement etc). In this way, the affected calls can be individually re-provisioned logically (over existing lightpaths), or physically (setting up a new-lightpath from ‘the source (s) to the destination (s) of the call), or by using a hybrid (physical and logical) approach. . . B.1 Conventional Lightpath Restoration ~ . ‘The affected lightpaths are stored in order of descending bandwidth allocated to them prior to the failure. The optical layer then t r i 4 to restore the lightpaths one-by-one, starting with the lightpath with the highest bandwidth allocation. In the case.of a call that had multi-lightpath servicing, all lightpaths must be restored for the call to be restored. Failure to restore a lightpath is, consequently, a failure to restore all the calls previously traversing it. B.2 Sub-Lumbda Restoration. Each OXC controller is maintaining a list of all the calls and their characteristics associated with every trunk. After a trunk failure, the affected.calls are grouped based on . ~ their corresponding (S-D) pairs. The different (S-0) pairs are stored in .order of descending bandwidth. The algorithm then tries to restore the.calls one-by-one starting from~the ones that belong to the (S-D) pair with the highest bandwidth using the servicing order logical-hybridphysical. IVPERFORMANCE EVALUATION The performance of the proposed approach is evaluated by simulating a mesh-based network consisting of 14 nodes and 21 bi-directional links. Adjacent nodes are connected through bi-directional physical links that consist of 4 fibers (two in each direction), where each fiber is assumed to have 4 wavelengths. The wavelength channel ’ capacity is assumed to be OC-48. The sub-lambda requests have bps demands that are normally distributed around 400 Mbps with a standard deviation of 200 Mbps, in multiples of SOMhps. The edge routers are assumed to have enough interfaces and process all the traffic that can potentially pass through them. We further assume a dynamic logical topology; i.e. lightpaths are torn-down whenever the last call utilizing them departs the network. C a m m m d v.. on om --