Key Technologies And Experimental System Of Elastic All Optical Network

With the rapid development of Internet services such as cloud computing and big data, global telecommunications infrastructure experiences the traffic boom and the resource convergence, which requires the backbone optical networks to be capable of providing more bandwidths and carrying more services. Traditional WDM-based rigid lightpaths would not meet such demands. With the emergence of flexible grids and software defined networking (SDN), and the development of OFDM and Nyquist WDM technologies, all optical switch networks evolve to be spectrum-variable and provide bandwidth on demands, and finally become the elastic all optical networks (EAONs) with the virtues of ultra-dense spectrum granularity and multi-domain heterogeneous interconnection.Due to the features of spectrum variable, bandwidth variable and modulation format variable, there are spectrum contiguity constraint and spectrum fragments in EAONs. Besides, there emerges new technologies such as spectrum splitting and optical concatenation in EAONs. These features destroy the independence between wavelength layers and the independence between lighpaths. As a result, the resource model, the networking mechanism and the service pattern of EAONs become quite different from these of traditional WDM networks.This paper focuses on the above problems. In view of the convergence trends of all optical networks, IP networks and data centers, this paper studied some key technologies in EAONs, and gained a series of innovative research achievements in the fields of all optical resource evaluation and optimization, heterogeneous synergistic networking mechanism, elastic service provisioning and so on. On the basis of these achievements, this paper designed and constructed an innovative experiment system of EAONs. The innovative research achievements are listed below.(1) For the dynamic performance evaluation problem of the spectrum resource in EAONs, this paper studied the dynamic features of flexible grids, and presented an exact performance analytical model which can exactly describe the stochastic features of the elastic spectrum allocation system. This paper studied the state representation of the spectrum allocation system, and proved that single dimensional random variables, which are widely used in the model of traditional WDM systems, cannot describe the stochastic features of spectrum contiguity constraint. Then, this paper defined a multi-dimensional Markov random variable based on the states of frequency units, which is capable of exactly describing the state of a spectrum allocation system. This paper studied the transition between system states, and found the method to evaluate the scale of state set. This paper studied the ergodicity, the reversibility and the reducibility of the model. Based on the characteristics of the transition intensity matrix, this paper presented the heuristic algorithms to solve the steady state distribution, and further presented the formulae of blocking probability, resource utilization rate and fragmentation rate, which can be used to evaluate the network performance and capacity. The analytical results match closely with the corresponding Monte Carlo simulation results, with the matching degree of 96.4%.(2) For the multi-layer topology resource mapping optimization problem of EAONs, this paper studied the mapping features of IP over EAON networks between service layer, IP layer and optical layer. Based on the spectrum splitting technique in EAONs, this paper proposed an electronic-optical-hybrid multi-flow Integer Linear Programming (ILP) model, which gives expressions to the spectrum contiguity constraint and distance-adaptive modulation format in EAONs. Besides, this ILP model reflects the physical features of each optical channel and each IP flow. Based on this model, this paper presents an energy efficient ILP model which gives expressions to the energy consumption of main optical devices and electronic devices. Furthermore, this paper also presents an energy efficient multi-flow service mapping heuristic algorithm for the same scenario. Simulation results shows that the energy efficient ILP model could reduce energy consumption by 19.55% on average, and the heuristic algorithm could reduce energy consumption by 14.1% on average. Besides, the study found that IP layer consumes almost 80% the sum of energy.(3) For the multi-domain heterogeneous synergistic networking problem, this paper studied the mechanism of setting up cross-domain path while keeping domain’s topology information isolating from others under the centralized control based on SDN. This paper modified the SDN architecture by setting up the orchestration plane over the control plane, and presented a collaborative multi-domain heterogeneous networking architecture. Based on this architecture, this paper presented 3 collaborative routing scheme, named as controller driven routing scheme (ConDRS), orchestrator driven routing scheme (CIDRS), and CIDRS with dynamic optimization (C1DRS-DO). This paper expanded the OpenFlow protocol and designed a JSON-based northbound API to support the circuit switched feature and the flexible grid feature of EAONs. Experimental results show that all the schemes could achieve cross-domain resource allocation efficiently. The signaling latency of ConDRS grows linearly with domain numbers, and the performance of C1DRS-DO is the best.(4) For the elastic service provisioning problem of EAONs, this paper designed and built an EAON innovative experiment system for virtual network resource provisioning. This paper studied the slicing and sharing problem of optical resource, and designed a novel network-function-virtualization-enabled optical note architecture with the programmable node control board and the data processing fabric as two newly designed functional devices. This paper conducted experiments of virtual all optical network provisioning and virtual network function provisioning. The experimental results show that the proposed networking architecture and virtualization scheme could reduce network cost by 20% on average and reduce blocking probability by 8% on average. Besides, the signaling latency could be kept within 100 ms.