orenais
commented
5 years ago There is no real absolute rules. It depends on the case you have to solve. You need an algorithm that can provide a path respecting some constraints. Typically you can use an RWA (Routing and wavelength assignment) algorithm which provides wavelength continuity, or if you have a flexgrid infrastructure an RSA (Routing and Slot Assignment) which provides slots continuity and contiguity. These algos shall support operators’ constraints such as robustness (taking into account SRLG and disjunction), latency (minimum distance), cost (limiting regens at the cost is some cases of a lower rate), spectral occupancy (limiting spectral use and fragmentation)… For the implementation of such algorithms, we usually use k-shortest-paths, but others can be used. As soon as several constraints must be considered, using a simple Dijkstra becomes difficult. GNPY as an example (an OpenSource impairment aware path calculation solution) uses Dijkstra for simple demands, and k-shortest path for more complex demands (include and exclude options) based on Python networkx library.
vijayanandc159
fgruman
EstherLerouzic
I would like to know what algorithm is recommended for path computation across WDM network in the PCE. Will it be a modified version of dijkstra - with one run to check for constraints at every hop and find shortest path by metric , or the K-th shortest path algorithm to continuously explore many paths and check optical constraints( such as wavelength continuity, ROADM directionality constraint etc) on each of the paths one by one
Please give your suggestions/experience.