Minimized spectrum resource consumption with rescaled failure probability constraint in flexible bandwidth optical networks

We investigate the problem of minimizing spectrum resource consumption for a set of connection requests in a flexible bandwidth optical network under the constraint that the failure probability of each connection must be below a specified threshold. To optimize the total frequency slots consumed, three schemes are proposed and evaluated: the rescaled failure-probability-aware algorithm (RFPA), the traffic cognition algorithm with rescaled failure probability (RFPTC), and an integer linear programming (ILP) model. We also introduce two traditional Dijkstra's algorithms with load balancing and spectrum assignment by first fit and traffic cognition (FF_DB and TC_DB) to compare with the proposed three schemes. For static traffic in a small network, the total frequency slots consumed by the RFPA and RFPTC algorithms will approach that of the optimal ILP solution as K increases, aswellastheaveragehops (AH). Furthermore, the average rescaled failure probabilities (RFPs) of the RFPA and RFPTC algorithms are much better than thoseof the ILP solutions. Inaddition, similarly, the results of the heuristic algorithms, FF_DB, TC_DB, RFPA, and RFPTC, in the large network have the same characteristics as in the small network. For dynamic traffic in a large network, the RFPTC algorithm reduces blocking probability and makes the best use of spectrum resources compared with the other schemes, which also reflects that discontinuous spectrum fragmentation is greatly reduced by the traffic cognition method. However, both RFPTC and RFPA result in much higher average RFP but perform fewer AH compared to TC_DB and FF_DB.