Towards Elastic and Fine-Granular Bandwidth Allocation in Spectrum-Sliced Optical Networks

Abstract

To overcome the inefficiency of the rigid frequency allocation in traditional wavelength division multiplexing (WDM) networks, the idea of slicing the optical spectrum for elastic and flexible bandwidth allocation has attracted significant interest recently. The resulting network, namely, the spectrum-sliced elastic optical path (SLICE) network, can facilitate both the super-wavelength and sub-wavelength traffic accommodation by allocating an appropriate number of sub-carriers. Compared to traditional wavelength routed WDM networks (WRNs), SLICE networks have the advantages of higher spectrum efficiency (through the elimination of spectrum gaps or guard-bands when possible) and better signal quality (by overcoming various impairments), thanks to the orthogonal frequency division multiplexing technology. To accommodate traffic demands in SLICE networks, the process of routing and spectrum allocation (RSA) has to be employed, which is different from and more challenging than the traditional routing and wavelength assignment problem in WRNs. In this work, we comprehensively study the RSA problem assuming the presence of known static or off-line traffic. We formally define the static RSA problem and show the NP-hardness of the optimal RSA problem. Integer linear programing models are then formulated to achieve different optimization goals in SLICE networks. We further analyze the lower/upper bound of the spectrum resources (i.e., sub-carriers) in SLICE networks with uniform traffic demands. To efficiently resolve the RSA problem in a large-scale network, we also propose two efficient algorithms, namely, the shortest path with maximum spectrum reuse algorithm, and the balanced load spectrum allocation algorithm, to minimize the required number of sub-carriers in a SLICE network. Our results show that the proposed algorithms can match the analysis and approximate the optimal solutions from the integer linear programing model.