11.1. Introduction
As each connection request is presented to an optical network, a route is calculated and a path is selected through the network from the source to the destination. Path selection in optical networks can be contrasted with the procedure used in IP networks. In IP networks, each router independently makes a forwarding decision based on the destination address of each packet. This independence of routing decision at each network node may result in routing loops (see Chapter 9). Thus, to avoid loops, all the routers in a network are required to have identical topology information and use the same algorithm for route computation. Routing loops are not an issue in connection-oriented networks, where connection routes are explicitly specified. An example is the usage of the explicit route object during LSP establishment under MPLS-TE (Chapter 7). Thus, there is a much weaker coupling between path selection and other routing functions such as topology/resource discovery in connection-oriented networks as compared with connectionless (packet) networks. This allows us to consider a wider range of route optimization criteria in connection-oriented networks.
Suppose that topology and resource information pertaining to an optical network are available from either a distributed routing protocol or a network management system. A range of criteria can then influence the selection of a connection path from a source to a destination. These include:
Desired connection bandwidth: This can be specified in two ways: as an exact value, or as a range, [bmin, bmax], where bmin is the minimum bandwidth required and bmax is the maximum required. In the latter case, it is expected that the maximum possible bandwidth between bmin and bmax would be allocated.
Time delay: This can be a constraint on the absolute delay along the path from the source to the destination, or a differential quantity with respect to another path or group of paths. This delay bound would apply both to propagation and processing delays. Note that as compared with packet networks, queuing delays do not play a role in optical networks.
Degradation: This constraint applies to path computation specifically in transparent optical networks (see Chapter 10). In these networks, attenuation along various routes is an issue. It may be required to minimize the degradation or keep it below some prescribed level.
Diversity: It may be required that the connection path is physically diverse from the path(s) of other connections, or avoid certain geographic areas entirely (see Chapter 10). Diversity constraints ensure that working and protection paths are not affected by the same failure(s).
Resource sharing: How the connection is routed can impact resource sharing in the network. For example, under mesh protection, working paths must be routed in such a manner as to maximize resource sharing among protection paths [Labourdette+02, Bouillet+02a] (see Chapter 10).
Network optimization: It may be desired that the overall network utilization is optimized. Thus, connection routing must be such that the use of network resources (capacity) is optimized or network costs are minimized [Labourdette+02].
It is clear from the above list that there are multiple dimensions to the problem of choosing a good route for a connection. Depending on the specific optical technology, some of these will be more important than others. For example, the Bit Error Rate (BER) in modern SONET/SDH networks is so low that signal degradation would not be a criterion for route selection. In transparent optical networks, however, there are a number of measures of degradation besides BER that have to be considered during route selection. In the following, we give an overview of route computation/path selection techniques and illustrate their applicability to optical networking.