A spectrum defragmentation problem in elastic optical networks was considered under the assumption that all connections can be realized in switching nodes. But this assumption is true only when the switching fabric has appropriate combinatorial properties. In this paper, we consider a defragmentation problem in one architecture of wavelength-spacewavelength switching fabrics. First, we discuss the requirements for this switching fabric, below which defragmentation does not always end with success. Then, we propose defragmentation algorithms and evaluate them by simulation. The results show that proposed algorithms can increase the number of connections realized in the switching fabric and reduce the loss probability.

The Elastic Optical Networks (EON) provide a solution to the massive demand for connections and extremely high data traffic with the Routing Modulation and Spectrum Assignment (RMSA) as a challenge. In previous RMSA research, there was a high blocking probability because the route to be passed by the K-SP method with a deep neural network approach used the First Fit policy, and the modulation problem was solved with Modulation Format Identification (MFI) or BPSK using Deep Reinforcement Learning. The issue might be apparent in spectrum assignment because of the influence of Advanced Reservation (AR) and Resource Periodic Arrangement (RPA), which is a decision block on a connection request path with both idle and active data traffic. The study’s limitation begins with determining the modulation of m = 1 and m = 4, followed by the placement of frequencies, namely 13 with a combination of standard block frequencies 41224–24412, so that the simulation results are less than 0.0199, due to the combination of block frequency slices with spectrum allocation rule techniques.

Elastic optical networking is a potential candidate to support dynamic traffic with heterogeneous data rates and variable bandwidth requirements with the support of the optical orthogonal frequency division multiplexing technology (OOFDM). During the dynamic network operation, lightpath arrives and departs frequently and the network status updates accordingly. Fixed routing and alternate routing algorithms do not tune according to the current network status which are computed offline. Therefore, offline algorithms greedily use resources with an objective to compute shortest possible paths and results in high blocking probability during dynamic network operation. In this paper, adaptive routing algorithms are proposed for shortest path routing as well as alternate path routing which make routing decision based on the maximum idle frequency slots (FS) available on different paths. The proposed algorithms select an underutilized path between different choices with maximum idle FS and efficiently avoids utilizing a congested path. The proposed routing algorithms are compared with offline routing algorithms as well as an existing adaptive routing algorithm in different network scenarios. It has been shown that the proposed algorithms efficiently improve network performance in terms of FS utilization and blocking probability during dynamic network operation.