| With the fast developing of cloud computing/storing and big data, data-center-driven next generation optical networks have been given a widely attention. To satisfy the large-bandwidth, low-latency and energy-efficiency transport network requirement, elastic optical networks which benefited by its flexible bandwidth allocation is becoming a hot topic in optical networks. Meanwhile, stimulated by data-center services, two kinds of convergence trends in the general optical transport network are prominent, vertically converge of multiple layers of the networking resource such as packet and circuit (optical channel) network, and horizontally converge of multiple stratums of heterogeneous resource such as computing/storage infrastructure and networking infrastructure. How to optimize resource utilization is a key challenge for two kinds of convergence.Our work is based on elastic optical networks by following two convergence trends. The service-driven resource optimization technologies in elastic optical networks are deeply studied. Some products have been achieved as follows:Firstly, for the problem of multi-layer traffic grooming under vertical convergence, we study the different grooming functions based on different elastic optical transponder technologies. Energy-minimized traffic grooming Integer Linear Programming (ILP) and two heuristic algorithms which are called Maximal Optical Layer Grooming (MOLG) and Maximal Electric Layer Grooming (MELG) are proposed. Simulation results show that significant energy saving is achievable using sliceable elastic optical transponder through minimizing the number of energy components. On the other hand, for dynamic scenario, we propose a novel auxiliary graph model with corresponding spectrum reservation schemes in sliceable-transponder equipped elastic optical networks. The edge weights in the auxiliary graph can be adjusted for different grooming purposes. Four dynamic grooming polices are compared and analyzed under two resource provisioning in terms of bandwidth blocking probability, number of transponder per request, number of physical and virtual hops per request.Secondly, for spectrum utilization of time-varying traffic under vertical convergence, a time dependent spectrum resource sharing model (T-SRS) motivated by cognitive radio networks is proposed. The idle spectrum can be reused by other users to improve the spectrum utilization. An interference factor (IF) is introduced to evaluate the performance of T-SRS model. We formulate the problem using ILP formulations to minimize IF optimally. Two efficient heuristic algorithms, which are called Maximum Holding time First with Shortest Path and Furthest Fit (MHF-SPFF) and Maximum Holding time First with Balanced-Interference and Furthest Fit (MHF-BIFF), are proposed and compared with ILP results. The numerical simulations show that more idle spectrum can be reused by reducing the IF. MHF-BIFF outperforms MHF-SPFF.Thirdly, for horizontal convergence, we study elastic optical networks virtualization for data-center network. A novel virtual resource description for elastic optical networks is proposed. We describe the network element by introducing link sliceability, transponder slice-ability and switch slice-ability. Two virtual networking embedding schemes based on link slicing (LS-based) and node slicing (NS-based) are proposed. Both of them are compared with Baseline scheme. Simulation results show that both LS-based and NS-based achieve better performance than baseline scheme in terms of blocking probability and revenue-to-cost ratio, NS outperforms LS.Lastly, for the control plane technology for both vertical and horizontal convergence, an OpenFlow-based control plane for elastic optical networks is proposed. The extension of OpenFlow protocol aims adjustment via extended OpenFlow protocol. Wireshark captures of the signaling procedure are printed out. Additionally, the overall latency including signaling and hardware for lightpath setup and adjustment is also reported. |
PDF Full Text Request