Protection and Restoration

There are basically two types of recovery, the difference being mainly the approach towards resource reservation, namelyProtection andRestoration.

If resources destined for recovery are allocated before the fault, for example at the time a particular connection is established, then the approach is called

“protection”. On the contrary, if resources are sought later on when the recovery procedure is in effect, the approach is called “restoration”. Within these two groups, further variations exists based on scope, capacity usage, path setup methods, and so on.

Recovery can be applied to LSPs, segments (a subsequence of links on a path), or links. If LSP protection is used, one or more backup paths are fully established to protect one or more working paths, implying that route computation was completed, the paths were fully signaled all the way, and that resources were allocated and cross-connected between the ingress and egress nodes. In essence, protection means that no signaling takes place to establish the backup path(s) when a failure occurs because it had already been done at setup.

Restoration means that some paths may be pre-computed, signaled, and selected a priori as backup, but not cross-connected. The complete establishment of the backup path occurs only after the working path fails, and requires additional signaling.

The advantage of protection over restoration is its fast recovery time and, in some of its variants, the guarantee that backup resources will be available when needed. On the other hand, restoration techniques can be more flexible with respect to which failure scenarios can be handled and at the same time its capacity requirements can be lower compared to protection [142].

Given the diversity of transmission technologies, topologies, and protec- tion strategies that GMPLS is called to support, a large number of recovery mechanisms have been proposed. Taxonomies and further references can be found for example in [24],[65],[37].

Types of Path Protection

Within the scope of protection, there are some types of recovery that are worth mentioning here, namelydedicated and shared path protection. Their

importance lies in the fact they are present in the all the major transport technologies.

As previously discussed, one drawback of protection compared to restora- tion is that it uses more capacity. One way to reduce this potentially wasteful consumption of resources is to break the one-to-one association between work- ing path and backup path. This basic idea gives rise to the following usual combinations:

1. 1+1 Type: Dedicated Protection, or DPP: One dedicated protection path protects exactly one working path, and the normal traffic is permanently duplicated at the ingress node on both the working and protection path. Both paths are link/node disjoint. The receiving end chooses which one is better. No extra traffic can be carried over the protection path.

2. 0:1 Type: Unprotected: No specific recovery path protects the working path. However, the working path can potentially be restored through any alternate available route/link, with or without any pre-computed restoration route. No resources are pre-established for this recovery type.

3. 1:1 Type: Dedicated Recovery with Extra Traffic: One specific recovery path protects exactly one specific working path, but the normal traffic is transmitted over only one LSP (working or backup) at a time. Both paths are link/node disjoint. Extra traffic can be transported using the recovery path resources.

4. 1:N Type: Shared Recovery with Extra Traffic: A specific recovery path is dedicated to the protection of up to N working paths (withN >1).

The set of working paths is explicitly identified. Extra traffic can be transported over the recovery path. All these paths must start and end at the same nodes.

Sometimes, the working paths are assumed to be resource disjoint in the network so that they do not share any failure probability, but this is not mandatory. If more than one working path in the set of N is affected by some failure(s) at the same time, the traffic on only one of these failed paths may be recovered over the recovery path. The choice of N is a policy decision. This type is applicable to both protection and restoration. This type of sharing is usually called Shared Path Protection (SPP).

2.4. NETWORK FAILURE AND RECOVERY

5. M:N Type: A set of M specific recovery paths protects a set of up to N specific working paths (withM, N >1, N ≥M). The two sets are explicitly identified. Extra traffic can be transported over the M recovery paths when available. All the paths must start and end at the same nodes.

Similar to the 1:N Type, sometimes the working paths are assumed to be resource disjoint in the network so that they do not share any failure probability, but this is not mandatory. If several working paths in the set of N are concurrently affected by some failure(s), the traffic on only M of these failed paths may be recovered. The choice of N and M is a policy decision. This type is applicable to both protection and restoration.

The Robustness of Complex Systems: A 3

Review of Measurements

This dissertation is concerned with the robustness of large communications networks, of which the Internet is the paradigmatic example. In fact, the Internet has been identified as an instance of what has been termed acomplex network, a subject that currently attracts tremendous interest and is being studied in connection with the so-called Network Science [87].

Although there is still no generally accepted definition of the essential characteristics of “complex networks”, some authors [13] propose that their main traits are as follows: First, they are not globally-engineered systems but rather the spontaneous outcome of the interactions among the self-organized units that constitute them. Second, the system behavior is an emergent property that cannot be understood by studying each system subunit in isolation. Third, they exhibit self-similarity, so that the fluctuations and heterogeneities observable in their structure span all scales of the system.

Andfourth, they are usually dynamic systems, showing complex, evolving topological features, such as hierarchies and communities. As daunting as complex networks may seem, judging them by their characteristics, the concepts and tools used to study them are shared with other disciplines, such as graph theory, physics, chaos theory and even epidemiology [87].

This chapter aims to summarize the main existing measures of robustness applicable to data networks, from the purely structural (i.e. topological) point of view as well from the perspective of system function. To that end, we start the chapter with a review of graph concepts and properties. Then follows an overview of the main network models that researchers have used to study telecommunication networks. Afterwards, we discuss a number of measures of robustness found in the literature and, finally, give a summary of the tools available for network topology generation.