• No results found

Survivability in optical networks

Providing resilience against network failures is an important requirement in network design today. A network connection, between a source to a destination, goes through several networking components (OLTs, OXCs, OADMs, fibers, routers etc., ). Each network component can fail during transmission. Examples of the causes of failures would be power outages, accidental cable cuts, or failures in electrical parts inside network elements. Network failures can be categorized into node failures (e.g., OXCs, OADMs, IP routers) and link failures (e.g., fiber-cables cuts and amplifiers). When a failure occurs, the backup mechanism establishes an alternative path to carry the interrupted traffic. If the alternative path is computed before the failure occurs, we refer the technique as protection. If it is computed after the failure occurs (i.e., dynamically), we called the backup mechanism as restoration [106]. Both the IP layer and the optical layer need to be resilient to failure. Restoration mechanisms are widely deployed at the IP layer, while the optical layer uses both kinds of backup mechanisms [52].

In order to address all failures without redundancy protection, in the context of a multi-layer recovery strategy, each layer (IP/optical) is responsible for providing protection against certain types of failures. The upper layer can provide the protection for failures in the lower layer if the lower layer can notify the upper layer about the failures.

If failures occur in IP routers, the recovery must be dealt with by the IP layer. This is a restoration technique since IP packets are routed over the failed nodes (i.e., routers) using the routing technology of the IP protocol. If failures occur in

the physical layer (e.g., fiber-cable cuts), either the IP layer or the optical layer is responsible for providing resilience. The optical layer can route the traffic of failed links over a predefined backup path. The protection at the IP layer is more flexible but slower than that at the optical layer.

2.3.1

Protection in the optical layer

Protection techniques at the optical layer can provide protection against several types of network failures such as single-link failures, single link/node failures, and multiple link failures. Most networks provide protection against single link failures. Some networks provide protection against node failures and multiple link failures for a given group of nodes/links, especially in the context of Shared Risk Link Groups (SRLG).

Protection techniques at the optical layer (i.e., the physical layer) require some physical redundancy within the network and protocols for rerouting traffic around the failure using this redundancy. One solution is to have a backup path for every working path. During normal operation, no traffic or low priority traffic is sent across the backup path. In case of failure, the higher-priority traffic will be sent over the backup path. The backup paths are computed before failure happens, thus it is called protection. To save network capacity reserved for protection, each backup link can be shared by multiple independent backup paths. Independence means that for a given failure, those backup paths sharing a link, will not be concurrently used. This is called shared protection.

Protection schemes can be categorized into three groups, based on the network structure they intend to protect: path-based schemes, link-based schemes, and segment- based schemes (Figure 2.7). In general, link-based schemes are faster (as only two end points of a failed link involve in the restoration process, the rest of the nodes on a working path can keep the same configuration) but path-based schemes use less bandwidth (since we use global information to choose a backup path with the cost almost as good as the working path).

(a) Link-based scheme

(c) Segment-based scheme (b) Path-based scheme

Figure 2.7: Protection schemes at the optical layer.

2.3.2

Protection at the logical layer

The protection at the optical layer, based on some physical redundancy within the network, is fast since we do not need to go up to the upper layer and do intensive signaling. If a failure is entirely in the physical layer, it can be handled by protection at the physical layer. That means, there is no need for protection at the logical layer. However, while protection at the optical layer is fast and easy to implement, it is costly. The traffic of an IP request is usually much smaller than the bandwidth of a wavelength, it would not be economical to use an entire wavelength to protect an IP request. Moreover, IP requests may have different QoS requirements, it is possible that some high priority IP requests need protection while others only require best effort services. Protection at the logical layer can help save cost by offering a more flexible protection scheme.

When protection in optical networks is not deployed, a network failure (e.g., power outages, cable cuts) can result in several logical broken links which share the same physical resource. Those logical broken links, in turn, can make the logical topol- ogy disconnected. The IP layer has the capability of rerouting traffic, i.e., resilient to faults if the network (i.e., the logical topology) remains connected. Hence, the necessary condition for the existence of a restoration scheme at the IP layer is that

the logical topology remains connected (survivable) with enough bandwidth in case of any network failures.

Related documents