Design For Survivable Elastic Optical Networks

Elastic optical networking (EON) has recently been regarded as a most promising building block for future optical networks. Comparing with the fixed-grid wavelength division multiplexing (WDM) networks that operate on discrete wavelength channels (50 or 100 GHz), EON can achieve agile spectrum allocation in the optical layer by using advanced technologies, e.g., optical orthogonal frequency division multiplexing (O-OFDM). On the other hand, due to the ultra high transmission rate in EON, a single component failure can cause tremendous data loss. Therefore, survivability becomes vital for EON.To overcome the drawback of low restoration speed or high employment cost from existing survivability designs, this article proposes to leverage the concept of failure-independent path-protecting preconfigured-cycle (FIPP-p-cycle) and design fast and cost-efficient protection schemes for protecting the lightpaths in EON against single link-failures. We fist study the problem of static network planning with FIPP-p-cycle protection, formulate an integer linear programming (ILP) model and prove its NP-hardness. To reduce the computing complexity, we design three time-efiicient heuristic algorithms, namely, protection-efficiency-based FIPP-p-cycle design (PE-FIPP), inte-grated FIPP-p-cycle design (Inte-FIPP) and maximum-independent-set based FIPP-p-cycle design (MIS-FIPP), for approximating the optimal solutions. This article also investigates dynamic FIPP-p-cycle configuration algorithms, and design a backup re-configuration scheme based on MIS-FIPP to improve the protection efficiency of FIPP-p-cycles during dynamic network operations. Simulation results indicate that the pro-posed FIPP-p-cycle algorithms can achieve better spectrum utilization and lower block-ing probability.Protection designs targeting for single link-failures do not consider the varies ser-vice availability requirements in real networks, e.g., a live video conferencing does not tolerate any disrution, while a file transmission only cares the average bandwidth it ac-quires. Therefore, a single protection strategy for all the services can not facilitate more efficient spectrum utilizations.In this article, by incorporating the flexible spectrum allocation and bandwidth-squeezed restoration schemes, we design a novel availability-aware service provision-ing framework for EON. We propose to use different path-based protection schemes, i.e., unprotected, dedicated-path protection (DPP) and shared-path protection (SPP), to satisfy different availability requirements, and develop the related theoretical model for availability analysis. On such basis, we design an availability-aware differentiated pro- tection (ADP) as well as an availability-aware backup reprovisioning (ABR) algotithm for EON, where ABR can intelligently adjust the backup resouce assignment according to the requests’evolving availability requirements for achieving more spectrum savings. To overcome the drawbacks from existing works, this article also develops a more accu-rate availability analysis model for p-cycle protection in EON by considering multiple link-failures and the relations between protection domains, and designs an availability-oriented dynamic p-cycle configuration algorithm accordingly. Simulation results in-dicate that the proposed analysis models can achieve better availability approximations while the availability-aware protection designs can effectively improve the spectrum efficiency and service availabilities.By seperating the control and data planes, software-defined networking (SDN) im-proves the network programmability and manageability. The combination of SDN and EON leads to software-defined elastic optical networking (SD-EON), which facilitates the agile optical-layer spectrum allocation in EONs. However, the control planes of SD-EON also encounter survivability issues, e.g., a controller failure can bring down the whole control plane, thus disrupting the service provisioning in the data plane.This article considers the scenarios in which controllers and control channels would fail simultaneously and studies the problem of survivable control plane establishment (SCPE) for SD-EON. Specifically, SCPE ensures that each node in SD-EON is equipped with at least two controllers (for primary and backup services). The controllers work in a mutual backup model, while the physical length of the control channel between each pair of controllers or controller and node should be lower than a preset threshold. We formulate an 1LP model for the SCPE problem, design a relevant-set based heuristic algorithm for solving it in large-scale topologies, and discuss the related system and protocol designs for enabling the deployment of SCPE based SD-EON. We build an OpenFlow based SD-EON testbed with SCPE, and experimental results show that the proposed framework can survive the SD-EON against varies kinds of data and con-trol plane failures, with also relatively good performances on control plane latency and scalability.