A prime-partition strategy for spectrum assignment problem in elastic optical networks

Abstract

Elastic optical networking is an attractive technology for the next generation of backbone networks, because it can flexibly provide spectrum resources to connection requests. In elastic optical networks, spectrum resources are generally split into frequency slices with finer granularity and each connection request can use several frequency slices according to its bandwidth requirement. In the process of spectrum assignment, numbers of frequency slices used by different connection requests are diverse, which may lead to misalignment of frequency slices in different links. The misalignment would affect the effective use of spectrum resources and result in the increase of blocking probability of connection requests. For the purpose of reducing the misalignment of frequency slices and thereby decreasing the blocking probability, we propose a prime-partition strategy for the spectrum assignment problem in this paper. The strategy partitions frequency slices in each link into prime blocks with sizes equaling prime numbers 2, 3, 5, 7, etc., and the prime blocks with the same size are aligned in different links. After the partition operation, each connection request can be served by one or more adjacent prime blocks with the same size, no matter how many frequency slices it needs. Based on the prime-partition strategy, we design a novel spectrum assignment algorithm in which the enhanced-most-used policy is used to break ties. The enhanced-most-used policy can make idle prime blocks more aligned in different links. Simulation results show that the prime-partition-based spectrum assignment algorithm outperforms the existing well-performed algorithms in terms of blocking probability.