The detailed dimensioning steps include:



Step 1: Disintegrate Demand Matrices

In general, as long as a layer is not the lowest transmission layer, it can have a native external demand matrix. Therefore, in a network probably there are multiple demand matrices that are in different network tributary levels. For example, in a network there can be an OC-3-tributary demand matrix in the OC-3 layer, and an OCH demand matrix in the Optical Channel (OCH) layer. Therefore, the dimensioning process is required to route all these demands independently layer by layer from top to bottom.

Step 2: Routing Client Layer Demands

For each demand in a client layer, Dijkstra's routing algorithm is applied to find the shortest path based on the topology in the bottom transmission layer. Depending on the protection type, a second protection path may also be found for each demand. In addition, demand splitting may be supported. As a result, the routing process will generate a set of routes, through which the traffic demands are routed.

Step 3: Bundling

After all traffic demands on a low-rate client layer are routed, they may be bundled in a server layer. For a bundling process, a client layer and a server layer that are related to the bundling process should be designated first. There are three types of bundling strategies, including (1) end-to-end bundling, (2) link-by-link bundling, and (3) bundling reoptimization.

End-to-end bundling

The end-to-end bundling strategy is simple enough to multiplex low-rate tributaries in the client layer into high-rate capacity trunks in the server layer in an end-to-end fashion. The bundling process will generate an intermediate demand matrix for the server layer. For example, if seven units of tributaries in the OC-1 layer are bundled onto the OC-3 layer, the bundling process will generate three units of OC-3 tributaries in the OC-3 layer, which is a demand in the OC-3 layer.

Link-by-link bundling

In contrast to end-to-end bundling, link-by-link bundling bundles traffic demand on a physical-topology-link or server-layer-topology-link basis. That is, all the traffic demands are routed based on the physical topology first, and then all the demands on a physical link or server-layer link are bundled. For example, if there are seven units of OC-1 routed over a common physical link, then on the serve layer (e.g., OC-3 layer) three units of OC-3 need to be established between the two end nodes of the server link.

Bundling Re-optimization

After bundling, a re-optimization process can be carried out to decrease the number of bundling links in the server layer. The purpose of the re-optimization is to clear traffic on some server links that have few traffic demands, by rerouting these demands onto other server links. If all the traffic demands are successfully re-routed, these server links can be removed from the design, as they are carrying no traffic demands. Different re-optimization strategies can be applied respectively for end-to-end and link-by-link bundling.

Step 4: Integrate Demand Matrices on an Intermediate Server Layer

From a bundling process, an intermediate server layer can have internal demand matrices generated from upper-layer traffic demands. Also, each network layer can have its own native demand matrix, which composes demands of customers who require network capacity in tributary of the current layer. These two types of demand matrices have the same level of tributary and make up a total demand matrix in the intermediate server layer. Thus, the integration of these two types of demand matrix will generate a total demand matrix for the current intermediate layer. This new demand matrix will be used for the following routing and bundling steps.

Step 5: Route Integrated Demand Matrix and Bundling

An intermediate server can also be a client layer of another lower server layer. The previous obtained integrated traffic demand matrix should also be routed in the physical layer and bundled on the next server layer. Again, Dijkstra's algorithm can be used to route the integrated demand matrix. Accordingly, a bundling and bundling re-optimization process can be performed to bundle the integrated traffic demands on the next server layer. The server layer is the current intermediate server layer. A similar demand matrix integration process can be further carried out to sum up an intermediate internal demand matrix and a native external demand matrix to form an integrated demand matrix for the current intermediate server layer. In summary, the above routing, bundling, and integration steps will be carried out recursively until the transmission layer. To that end, the dimensioning process is completed.

Step 6: Cost Evaluation, Equipment List, and Performance Analysis

Based on the previous network dimensioning, we can eventually make a total cost evaluation, list out all the equipment requirement and configuration, and make ?what-if? analyses for the new design such as reliability and availability analyses.