13. Scaling up PCE architecture

The introduction of a centralized function automatically leads to consideration of the consequences of scaling up. Indeed, if we consider a network of 100 nodes, where the PCE performs all the calculations, it will have to take over the calculation load initially distributed over the 100 nodes of the network. In addition to this computational load, the PCE will have to maintain the PCEP protocol sessions to the 100 nodes in the network, as well as the load linked to the protocol exchanges on these PCEP sessions. Finally, as the PCE opens the door to more complex calculations, the time required for each calculation could be greater.

On the face of it, it would seem that centralized PCE architecture would be difficult to scale up by addressing these points. However, it's important to understand that a PCE is different from a network device. The PCE is software (not...