In communication networks, routing involves the identification of a path between the source and destination nodes for each connection request. Traditional routing approaches find a path that minimizes a certain cost parameter, such as the length of the connection. Most of the reported routing and wavelength assignment algorithms (RWAs) assume that once an available path and wavelength have been identified, the connection is feasible. This may not be true in transparent optical networks using wavelength division multiplexing (WDM), where the optical signal experiences and accumulates the effects of physical impairments associated with the transmission line and optical switching nodes. In some cases, this results in unacceptable signal quality.
Hence, impairment-constraint-based routing (ICBR) is needed in order to ensure that the connections are feasible. To do this, it is necessary to consider not only the network-level conditions but also the equally important physical performance of the connection.
Previously we proposed an RWA that included an ICBR approach. Our technique took into account various physical impairments present in an optical network, and considered different problems independently. Here, a path was assumed to be feasible if a set of criteria reflecting the signal quality (in terms of the different impairments) is satisfied.1
Here we propose a scheme that integrates routing and wavelength assignment and also considers the physical performance of the network in an ICBR algorithm developed to achieve optimal network performance. This scheme takes into account—through detailed modeling—the various hardware issues and their interplay. This integrated approach produces an estimation of the signal quality (Q) factor and the associated penalties.
We use the example of a metropolitan area network (MAN): in a typical network of this type, amplifier spans are short and the per-channel launch power can be relatively low (without incurring notable penalties from noise), so nonlinear effects do not have a significant impact. We developed an analytical model of the linear physical impairments: these included chromatic dispersion (CD), polarization mode dispersion (PMD), amplifier spontaneous emission (ASE) noise, crosstalk, and filter concatenation effects. This model is based on Q-factor performance of the signals traversing the metro network and is given in the following equation:2
where Qout and Qin are the Q-factor values at the destination and source of each connection. The eye-related penalty is due to effects such as PMD, CD, and filter concatenation, while the noise-related penalty is due to effects such as ASE noise and crosstalk.
The proposed ICBR scheme has three phases. The first is a pre-processing phase where all the information related to the network characteristics and the traffic demands is collected. This includes network topology, equipment, infrastructure, link capacity, and fiber characteristics. The traffic characteristics include information such as number of demands, end nodes, bit-rate, etc. Based on these parameters, we compute the link costs that should be taken into account when finding the possible paths.
Once the link costs are computed, the second phase assigns paths and wavelengths to all the demands (a conventional RWA module may be used for this phase, e.g. shortest path). We identify a set of k shortest paths for each connection request, and solve a linear optimization problem to minimize the cost induced by all flows in the network.
The third phase checks the validity of the lightpaths delivered by the second phase. The performance factor Q, which is analytically modeled to include all of the physical defects present in the network, is used as the link cost parameter. If the proposed lightpath satisfies the signal quality requirements and takes into account all these impairments, then a path is established. If it doesn't, then the path is deleted from the set of k shortest paths, and the RWA phase is called again. If none of the paths satisfy the impairment constraints, then the connection is blocked.
Figure 1 illustrates the improvements that are possible in the blocking probability of a network when using ICBR instead of the conventional shortest-path RWA. In this graph, the network blocking probability has been plotted as a function of the amplifier span length, which is assumed to be uniform across the network. When applying Q-based routing, the overall network performance is improved, especially in terms of signal quality, which has a direct impact on improving the network blocking probability.
Figure 1. An impairment-constraint-based routing (ICBR) algorithm optimizes and improves overall network performance with a lower percentage of blocked calls as compared to the performance achieved in a traditional shortest-path (SP) approach.