Make-before-break based spectrum defragmentation in mixed-grid optical networks

Abstract

With flexible spectrum allocation and distance-adaptive modulation, the flex-grid elastic optical network (EON) has been seen as a promising solution for better supporting the ever-growing, heterogeneous, high-bit-rate, and high-volume traffic in optical backbone networks. A gradual migration from fixed-grid to flex-grid is preferred for network operators in order to decrease upfront investment, wastage of previously deployed devices, and disruptions to network regular operations. However, during the brownfield migration process, fixed-grid and flex-grid equipment coexist in the network, resulting in a mixed-grid optical network. Because of the interoperability constraints between fixed-grid and flex-grid technologies in a mixed-grid optical network, reducing spectrum fragmentation is more challenging than in an EON. In order to address this issue, we investigate the spectrum defragmentation in a mixed-grid optical network and propose a make-before-break based spectrum defragmentation (MBBSD) algorithm. Specifically, we first design a joint evaluation metric to prioritize requests for reconfiguration, with the purpose of selecting the most critical requests for reconfiguration to avoid excessive operational complexity. Then, based on the make-before-break (MBB) method, we propose an efficient reconfiguration algorithm to minimize disruption during spectrum defragmentation. We formulate the resource reallocation problem for the reconfigured requests as an integer linear programming (ILP) model and prove it to be NP-Complete. Furthermore, a heuristic algorithm for request reconfiguration in a large-scale mixed-grid optical network is developed, where the ILP model is not tractable. We conduct comprehensive simulations on various network scenarios to evaluate our proposals, and the results demonstrate that our proposed algorithm can reduce the bandwidth blocking ratio by 27.5%-56.4% for the traffic load of 800 Erlang in NSFNet compared to the other three benchmark algorithms. The proposed algorithm can also achieve 4.1%-18.6% reduction for spectrum utilization ratio for same load in NSFNet. Moreover, we can also reduce the highest used frequency slot (FS) index compared to the other three benchmark algorithms.