We consider the problem of dynamic connection establishment and spectrum defragmentation in flexible optical networks. When the spectrum is fragmented, blocking a connection establishment, the algorithm reactively re-optimizes the network by shifting (“pushing”) in the spectrum domain and/or rerouting existing connections. We start by presenting an algorithm based on integer linear programming formulation that searches among all combinations of shiftings and reroutings and selects the one that minimizes the changes in existing connections. We also present a heuristic algorithm that recursively shifts/reroutes connections around a void. The solution space of the heuristic can also be very large, so we use a threshold on the recursion depth to reduce the complexity and also provide a trade-off between performance and running time. Our simulation results show that the blocking probability can be substantially reduced using the proposed techniques as opposed to a network that does not reactively defragments the spectrum. The proposed heuristic achieves near-optimal performance, for cases that we were able to find optimal solutions, while the selection of the recursion threshold was shown to provide a good trade-off of performance for running time.
- Dynamic Routing and Spectrum Allocation
- Flexible/elastic optical networks
- Recursive process
- Spectrum defragmentation