Network Optimization And Controlling Mechanism In Spectrum-Sliced Flexible Optical Networks

The ever increasing traffic demand and various types of services determine the optical network as the underlying infrastructure to be more flexible and intelligent. On one hand, big-data services such as video service and cloud-type applications, require optical network for more transport capacity to accommodate those tremendous traffics. On the other hand, ever booming data-center networks demand optical network to be more dynamic and flexible, to satisfy various complex application demands. Therefore, how to optimize optical physical layer spectrum resources and how to improve the control mechanism, to improve optical network’s flexibility and efficiency, has become a major research issue concerning optical networking technologies.Based on spectrum allocation optimization and control mechanism in spectrum-sliced elastic optical path networks (SLICE), this doctoral dissertation focuses on several aspects such as spectrum allocation algorithms, flexible control plane, and realization of software defined optical networks (SDON), including several innovative research results as follows:(1) As for network optimization algorithm, asymmetric spectrum assignment (ASA) algorithm is proposed to improve network performance on physical layer. The key principle of ASA is that during spectrum allocation process, the spectrum resources on each direction should be assigned individually according to the respective bandwidth demand, which are more likely to be asymmetric. For the direction of less traffic demand, spectrum resources can be reserved to accommodate other services. Comparing with symmetric spectrum assignment which is adopted by conventional optical core networks, ASA is more flexible in term of spectrum allocation and is easier to fit application requirement from upper layer. This dissertation uses combination of multiple simulation scenarios and calculation modules, selects network blocking probability and spectrum utilization ratio as metrics of network performance, and constructs simulation platform based on SLICE network. The simulation results show that, comparing with symmetric spectrum assignment method used by traditional optical networks, the ASA algorithm can significantly reduce blocking probability and improve spectrum utilization ratio by appropriate spectrum allocation mechanism, therefore optical network performance is enhanced from physical layer.(2) As for network simulation, blocking probability balance (BPB) method is proposed to smooth blocking probability (BP) value distribution. In traditional network simulation, network BP value is one classic metric to evaluate network’s performance. However this BP value is the average of all nodes’individual BP value, and the BP value distribution among all nodes is ignored. To cope with that, BPB method is proposed in this dissertation. The key principle of BPB is that based on feedback mechanism in software defined networks (SDN), each node’s BP value is collected in real time and sent to controller, and those nodes whose BP values have exceeded the pre-set threshold should commence K shortest path (KSP) calculation, thus BP values between different nodes are equilibrated. The simulation is based on routing and spectrum assignment (RSA) problem under dynamic service demands in SLICE network. The simulation results shows that BPB has several advantages such as:comparing with traditional RSA process without KSP calculation, BPB alleviates BP value’s variation. Comparing with full KSP calculation without feedback mechanism, BPB reduces computation complexity. Therefore the BPB method achieves good balance among network BP value, node BP value distribution and computation effort, and takes both network efficiency and fairness into consideration. (3) As for networking experiment, a number of schemes based on software defined optical networks (SDON) are studied, which include following two parts:(a) NOX controller and extended OpenFlow protocol are adopted. With the vertical coupling of "Network manager-NOX controller-OpenFlow Agent (OA)-optical infrastructure", the test-bed aims to emulate data center network’s centralized control mechanism over optical nodes from different domains. Comparing with conventional ASON/GMPLS technology, the SDON-based test-bed reduces cross domain communication delay, and enables the fast lightpath provisioning process to be less than Is. In addition, the protocol complexity is reduced since only one OpenFlow protocol is needed, thus the development process of control plane is simplified,(b) Optical-layer service differentiation is studied. Commercial OpenFlow switches are utilized as edge routers to implement mapping between services of different priorities and different egress optical ports. The hybrid control plane comprises POX controller and RSVP-based network manager to control OpenFlow switches and optical nodes respectively. Optical-layer services differentiation is experimentally demonstrated. Comparing with traditional QoS technologies based on packet switching in electronic domain, optical-layer service differentiation enables physical separation of different services, and offers better security and bigger bandwidth. In addition, traffic load-balancing, and lightpath protecting and restoration of fiber level and wavelength level, are demonstrated to enhance optical network flexibility and survivability.