PaperlayerView exact revision on arXiv

Source-linked AI summary

Software Defined Optical Networks (SDONs): A Comprehensive Survey

Akhilesh Thyagaturu, Anu Mercian, Michael P. McGarry, Martin Reisslein, Wolfgang Kellerer

arXiv:1511.04376v3cs.NI

TL;DR

SDON research must address heterogeneous optical components, interfaces, and application-dependent service requirements. This paper surveys SDON infrastructure, control, applications, virtualization, and orchestration studies, identifying investigated mechanisms and future research directions.

  • Problem

    Heterogeneous optical components, numerous physical- and transport-layer interfaces, and differing application QoS requirements complicate SDON management and service deployment.

  • Method

    The paper comprehensively surveys SDON studies, organizing them around infrastructure, control, applications, virtualization, and orchestration.

  • Results

    The survey finds investigated approaches for flexible optical switching, performance monitoring, control frameworks, virtual optical network provisioning, QoS, security, energy efficiency, failure recovery, and multilayer or multidomain orchestration.

  • Takeaways & Limitations

    SDON studies have produced insights and demonstrated advantages for optical transport network management, while future work should improve integration, interfaces, standardization, and automation.

  • Takeaways & Limitations

    Virtual network embedding research has explored specialized objectives such as survivability, adaptability, and energy efficiency but lacks a comprehensive weighted framework combining them.

Abstract

from arXiv · show

The emerging Software Defined Networking (SDN) paradigm separates the data plane from the control plane and centralizes network control in an SDN controller. Applications interact with controllers to implement network services, such as network transport with Quality of Service (QoS). SDN facilitates the virtualization of network functions so that multiple virtual networks can operate over a given installed physical network infrastructure. Due to the specific characteristics of optical (photonic) communication components and the high optical transmission capacities, SDN based optical networking poses particular challenges, but holds also great potential. In this article, we comprehensively survey studies that examine the SDN paradigm in optical networks; in brief, we survey the area of Software Defined Optical Networks (SDONs). We mainly organize the SDON studies into studies focused on the infrastructure layer, the control layer, and the application layer. Moreover, we cover SDON studies focused on network virtualization, as well as SDON studies focused on the orchestration of multilayer and multidomain networking. Based on the survey, we identify open challenges for SDONs and outline future directions.

I. INTRODUCTION

SDN separates and centralizes network control through programmable abstraction layers, enabling flexible management and virtualization. This survey reviews SDON research across architectural layers and related optical networking areas.

  • SDN foundations: SDN separates control-plane and data-plane functions, centralizes control logically, and makes network functions programmable.Centralization broadens the controller’s network perspective while introducing scalability tradeoffs.
  • SDN architecture: The SDN architecture comprises infrastructure, control, and application layers connected through standardized interfaces such as the SBI and NBI.The control layer abstracts infrastructure capabilities for applications.
  • Optical networking motivation: Optical networks offer high transmission capacity but require specialized control because photonic transmission and switching operate across circuit, burst, and packet granularities.These characteristics motivate applying SDN flexibility to optical infrastructures.
  • Related work: The survey situates SDON research alongside prior reviews covering general SDN, flexible optical networking, access networks, and wireless networking.Earlier work included limited SDON sampling or focused on particular technologies and domains.
  • Survey scope: The survey provides a comprehensive academic review of SDON research organized by infrastructure, control, and application layers, unlike earlier narrower reviews.It also covers virtualization and orchestration across multilayer and multidomain networks.

2) Control Layer:

The SDN control layer programs network elements through southbound interfaces and exposes abstracted network state to applications. Controllers and orchestrators coordinate increasingly complex control decisions through multiple interface and protocol choices.

  • Control-layer roles: The SDN controller configures network elements through SBIs and can request flow statistics, topology, neighbor, and link-status information.Applications provide requirements while the controller translates them into infrastructure actions.
  • Orchestration: An SDN orchestrator coordinates software modules, controller hierarchies, or multiple controllers above the control and application layers.Unlike controllers that signal predefined rules, orchestrators make broader coordination decisions.
  • Control versus orchestration: SDN control and orchestration differ in scope: controllers implement specific functions such as routing, whereas orchestrators coordinate broader decisions across layers or domains.The distinction becomes important as SDN systems grow more complex.
  • Northbound interface: The NBI connects controllers with applications, while REST APIs support heterogeneous applications operating over a common controller.REST principles include client-server interaction, statelessness, caching, uniform interfaces, and layered systems.
  • Southbound interface: The SBI connects controllers with infrastructure devices through protocols including OpenFlow, PCEP, NETCONF, and BGP-LS.These protocols support flow control, path computation, device configuration, and topology-information collection.

C. Network Virtualization

Network virtualization abstracts optical infrastructure so multiple isolated virtual networks can share physical resources. SDON infrastructure studies address flexible photonic transmission, switching, monitoring, and network tiers supporting virtualization.

  • Virtualization principles: Network virtualization creates independent virtual network slices with prescribed resource allocations over one or multiple physical infrastructures.Hypervisors abstract network elements, links, and control functions into logically isolated slices.
  • Optical constraints: Optical-network virtualization must account for photonic transmission and switching characteristics across circuit, packet, and burst switching.Burst switching rapidly establishes short-lived optical circuits for coalesced packets, while packet switching requires optical forwarding and buffering mechanisms.
  • Network tiers: Optical networks are organized into access, metropolitan, and backbone tiers with distinct structures and connectivity roles.Access networks commonly use tree-like PONs, metro networks commonly use rings or stars, and backbone networks commonly use meshes.
  • Infrastructure studies: SDN-controlled photonic infrastructure studies examine configurable transceivers, space-division multiplexing, optical switching, and cognitive monitoring.Monitoring optical signal quality can dynamically control transceivers and switching-element filters.
  • Study classification: The infrastructure-layer classification includes bandwidth-variable transceivers, sliceable multiflow transceivers, space-division multiplexing, and optical switching topics.The classification organizes physical-infrastructure SDON studies by these component and technology areas.

1) Single-Flow Bandwidth Variable Transceivers (BVTs):

SDN-controlled BVT research makes optical transceivers programmable for flexible modulation, bandwidth, and signal-flow operation across mesh, PON, and data-center networks.

  • Single-Flow BVTs: Single-flow BVTs make photonic transmission characteristics of one optical signal programmable for flexible WDM networking.Related designs target general optical mesh, PON, and data-center networks.
  • Single-Flow BVTs: A software-defined coherent transponder more than doubles average transmission capacity per user versus a static approach in TDMA PON access networks.The flexible transponder digitally processes burst transmissions according to user distance from the OLT.
  • Single-Flow BVTs: DSP reconfigurable ONU and OLT designs coordinate shaping and matching filters for SDN-controlled optical OFDM communication in PONs.The OLT controller matches ONU filtering and maintains filter orthogonality.
  • Single-Flow BVTs: Around 3.7 Gb/s per ONU is achieved for eight upstream ONUs over a 25 km PON.This is the reported performance evaluation of the DSP reconfigurable ONU and OLT system.
  • Sliceable Multi-Flow BVTs: Multi-flow transceivers generate parallel optical flows and provide an infrastructure basis for optical network virtualization.Sliceable designs configure subcarrier modulation, bit rate, and power, while V-BVTs compose slices from independent subcarrier and modulator pools.
  • Sliceable Multi-Flow BVTs: Optimized sliceable-BVT transmissions can significantly reduce network cost and power consumption.The cited work reports this benefit for the associated hybrid long-reach fiber access architecture.

B. Space Division Multiplexing (SDM)-SDN

SDON infrastructure studies apply programmability to spatial multiplexing, switching elements, access aggregation, and hybrid packet-circuit architectures.

  • Space Division Multiplexing: SDN-controlled SDM enables bandwidth-flexible, programmable optical networks with network slicing and sliceable superchannels.Superchannels combine multiple spatial carriers to support dynamic bandwidth and QoS provisioning.
  • Space Division Multiplexing: A reconfigurable SDM radio access network connects centralized BBUs and distributed RRHs through multicore fiber.The architecture applies SDN-controlled switching and radio-over-fiber operation.
  • Switching Elements: Colorless, directionless ROADMs permit wavelength-channel and route changes through a management control plane, unlike statically configured networks.The flexibility addresses operational costs associated with physical intervention in conventional ROADM networks.
  • Switching Elements: An OXC-backplane CDC ROADM achieves higher flexibility and lower losses than prior wavelength-selective- and multicast-switch designs.The comparison is analytical, with experimental evaluation of the backplane-based ROADM also reported.
  • Switching Elements: An optical white box combines programmable optical and electronic backplanes to support softwarized switching and network functions.A prototype used a 192 × 192 optical space switch, and virtual switching-node creation took roughly 400 ms.
  • Switching Paradigms: Hybrid packet-circuit optical networking uses OpenFlow extensions and common flow abstractions to unify control of both switching paradigms.The approach is described for core networks and extends matching rules and flow entries for hybrid operation.
  • Switching Paradigms: For an intra-PON traffic ratio of 0.3, QPAR achieves power savings up to 24% by optically rerouting traffic between OFNUs.The rerouting avoids passage through OLFTs and supports communication between wireless base stations.

D. Optical Performance Monitoring

SDON performance-monitoring and control studies combine sensing, programmable optical elements, and global optimization to adapt transmission conditions and improve signal quality.

  • Optical Performance Monitoring: CHRON combines monitoring elements, software-adaptable elements, and cognitive processes to adapt optical-network operation to current conditions.It monitors non-catastrophic and catastrophic optical transmission impairments.
  • Optical Performance Monitoring: SDN provides the framework for collecting monitoring information, executing control algorithms, and instructing adaptable transceiver and wavelength-switch/amplifier components.The controller coordinates sensing and control decisions across these components.
  • Optical Performance Monitoring: Joint EDFA gain and filter equalization control targets OSNR degradation caused by amplifier operating points and accumulated filtering effects.The studies exploit the SDN controller’s global perspective to coordinate these controls with wavelength assignment.
  • Optical Performance Monitoring: Close to 5 dB OSNR improvement is achieved in a four-ROADM testbed with 100 km fiber links through global EDFA and filter optimization.A separate 14-node simulation found mutual interactions among EDFA gain, filter equalization, and wavelength assignment.
  • Summary and Discussion: SDON infrastructure research has produced programmable transceivers and switching elements while indicating potential for lower costs, flexible resource management, higher utilization, and lower latency.The survey identifies comprehensive component-interaction optimization and broader hybrid-network support as future research directions.

3) Controlling Optical Packet and Burst Switches with OpenFlow:

OpenFlow can be extended or abstracted to control optical packet and burst switches, including devices that lack native SDN interfaces or optical buffering.

  • Controlling Optical Packet and Burst Switches: OpenFlow flow tables can express optical packet-switch forwarding rules while offloading their computation to an SDN controller.This can simplify the design of complex optical packet switches.
  • Controlling Optical Packet and Burst Switches: OPS control can use an abstraction layer, electrical handling of unmatched packets, and wavelength identifiers added to flow-table entries.The abstraction layer converts OpenFlow messages into native OPS configuration.
  • Controlling Optical Packet and Burst Switches: A demonstration system combines SDN-controlled optical packet switching with WSS-based optical circuit switching and an SOA switch.The packet portion is SDN controlled, while circuit switching uses wavelength-selective switches.
  • Controlling Optical Packet and Burst Switches: Optical burst switches can send a Burst Header Packet to the SDN controller when no flow-table entry exists for a burst.This lets the controller process a new flow using the burst header rather than the first packet.
  • Retro-fitting Devices: A hardware abstraction layer retrofits non-SDN optical devices for OpenFlow control by translating controller messages into native interfaces such as SNMP, TL1, or proprietary APIs.Virtual OpenFlow ports correspond to physical device ports, allowing flow entries to be mapped between them.
  • Retro-fitting Devices: A non-SDN PON OLT can be retrofitted with a two-port OpenFlow switch and abstraction layer so OpenFlow controls its switching functions.The layer translates OpenFlow forwarding rules into commands understood by the OLT.

C. SDN Control of Optical Network Operation

SDN control applies centralized, programmable management to optical access, tandem, and elastic networks, while optical timescales and fragmentation constrain controller-based operation. Surveyed studies report gains from global coordination, spectrum defragmentation, and application-specific bandwidth allocation.

  • Controlling Passive Optical Networks with OpenFlow: SDN-controlled PONs can use upgraded SDN-OLTs or retrofit non-SDN OLTs with OpenFlow switches and a hardware abstraction layer.OpenFlowPLUS extends programmability to OLT and ONU devices and adds PON-specific flow actions.
  • Controlling Passive Optical Networks with OpenFlow: Many OLT functions are too latency-sensitive for centralized control, but ONU registration, coarse-timescale DBA, and power control can be offloaded effectively.The limiting factor is latency between the controller and OLTs.
  • Elastic Optical Networks: Optical spectrum fragmentation can leave sufficient capacity unusable because demand requires adjacent spectrum, motivating periodic defragmentation.SDN controllers can use their broad network perspective to make defragmentation more effective; reported blocking-probability reductions range from 3% to 75%.
  • SDN Control of Tandem Networks: 40% increase in network bandwidth utilization resulted from coordinating access- and metro-stage SDN controllers across several physical PONs.The comparison is against bandwidth allocation performed only within individual PONs.
  • SDN Control of Tandem Networks: SDN applications support coordinated bandwidth allocation and routing across access, metro, and core segments, including VM migration and IoT traffic.For IoT, multicast and flow suspension prioritize urgent traffic over routine flows.
  • SDN Control of Tandem Networks: 30–50 % improvement was reported for request serving ratio, revenue, and delays using SDN-based multicast and flow suspension in a virtualized WDM access network.The results came from a real-time SDN controller implementation.

2) Path Computation Element (PCE):

PCEs separate path computation from forwarding, and hybrid GMPLS/PCE–SDN designs combine centralized computation or control with different southbound mechanisms. Comparative studies report trade-offs among blocking, utilization, setup time, scalability, and controller capacity.

  • Path Computation Element (PCE): A PCE computes network paths from topology and criteria, optionally using bandwidth, optimization metrics, and link constraints supplied in a request.Replies return an explicit route or indicate that no path exists.
  • Path Computation Element (PCE): A stateful PCE maintains topology and provisioned-lightpath databases, computes an optimal path, and configures GMPLS optical nodes through NETCONF or proprietary CLIs.This combines path computation with instantiation capabilities.
  • Hybrid GMPLS/PCE and SDN Control: Hybrid control lets an SDN controller use a PCE to control part of the infrastructure via PCEP or use PCE path computation while configuring elements through OpenFlow.These are the two architectures illustrated in Fig. 9.
  • SDN vs. GMPLS: OpenFlow produced higher wavelength utilization and shorter average lightpath setup time, while GMPLS produced slightly lower blocking probability in a 1000-node comparison.The comparison evaluated blocking probability, wavelength utilization, and lightpath setup time.
  • SDN vs. GMPLS: A full-SDN network enabled path lengths approximately one third of those computed for a non-SDN network.The study links shorter paths with reduced energy consumption.
  • SDN vs. GMPLS: 4 % reduction in lightpath blocking probability and nearly half the lightpath setup time were reported for SDN OpenFlow compared with GMPLS.An Open vSwitch-based controller also processed about three times as many flows per second as a NOX-based controller.
  • SDN Controller Flow Setup: The length of short-lived “mice” flows was the dominating parameter in the evaluated controller stability model.The model varied mice-flow length, elephant-flow fraction, and elephant-flow traffic volume.
  • SDN Controller Flow Setup: With 20 kB mice flows and 1-second circuit lifetimes, increasing packet traffic from 0.1 to 0.9 reduced supported network size from 14 to 5 nodes.The corresponding wavelength count decreased from 92 wavelengths per node to 10 for a single Beacon controller.

3) Out of Band Control:

The survey covers out-of-band and distributed control architectures, clustering, and virtualization across access, metro, core, and data-center networks. These approaches address control reliability, scalability, and resource isolation, but controller-performance evidence and virtualization scope remain limited.

  • Out of Band Control: Out-of-band control separates SDN control traffic from data traffic, increasing control reliability during ring-network failures at the cost of a separate physical network.The cost is especially high for a single-wavelength ring.
  • Distributed and Clustered Control: Clustering partitions wavelength-switched networks among SDN controllers while preserving global topology views and inter-cluster lightpath establishment.Controllers request establishment within other clusters through a westbound interface when paths cross cluster boundaries.
  • Distributed and Clustered Control: Lightpath establishment time decreases exponentially as the number of control clusters increases.Clusters can be formed to minimize the average distance to each SDN controller.
  • Scope of SDON Control Studies: The literature includes OpenFlow extensions, non-SDN retrofits, GMPLS/PCE integration, and centralized-control use cases such as large-scale bandwidth allocation.These studies span optical circuit switching, optical packet switching, and passive optical networks.
  • Control Layer: Summary and Discussion: SDON controller performance analysis remains immature, with insufficient evidence linking controller implementation and network characteristics to objectives such as maximum flow setup time.This limits informed partitioning of networks into control domains.
  • Classification of SDON Virtualization Studies: Optical virtualization research spans access, data-center, metro/core, virtual optical-network embedding, hypervisors, and cloudnets.The classification includes impairment-aware, flexi-grid, survivable, dynamic, and energy-efficient embedding studies.
  • OFDMA Based PON Access Network Virtualization: Distinct OFDMA subcarriers isolate VPON slices, allowing different MAC standards to operate over the same physical PON infrastructure.A central SDN control module manages OFDMA transmissions, receptions, and virtual MAC processing.

2) FiWi Access Network Virtualization:

The survey covers virtualization across FiWi access, cloudnet, and metro/core optical networks, including embedding strategies for survivability, impairment awareness, and resource optimization.

  • FiWi Access Network Virtualization: Virtualized FiWi networks provide a unified view across heterogeneous fiber and wireless segments, simplifying flow control and operational algorithms.Testbed measurements report improvements in nodal throughput, link bandwidth utilization, packet delay, and video-streaming QoE.
  • FiWi Access Network Virtualization: WiMAX-VPON uses a common MAC protocol across wireless and fiber segments with VPN-based admission control and bandwidth allocation for per-flow QoS.Discrete-event simulations demonstrate favorable performance.
  • Optical Data-Center Networks: LIGHTNESS targets dynamic, programmable, and highly available data-center connectivity through SDN-controlled optical packet, circuit, and hybrid switches.
  • Metro/Core Virtual Network Embedding: Virtual optical network embedding maps virtual nodes and links onto physical optical substrates while addressing survivability, impairment, WDM, and flexi-grid constraints.Studies formulate optimization problems and develop heuristics because some embedding formulations are NP-hard.
  • Metro/Core Virtual Network Embedding: 5–20% lower blocking probabilities are reported for parallel survivable embedding compared with a sequential benchmark, but protection covers only a single link or node failure.Planning and integrated cooperation can also improve revenue and request acceptance ratio.

2) Hypervisors for VONs:

Hypervisors expose physical optical infrastructure as isolated virtual optical networks and enable controllers to manage slices and virtual resources. The surveyed application-layer studies use SDN controllers for QoS-oriented services and traffic adaptation.

  • Hypervisors for VONs: An optical-network hypervisor presents one physical substrate as multiple isolated VONs and translates VON-controller actions into physical-network control.
  • Hypervisors for VONs: T-NOS implements a virtualized GMPLS control plane with offloading for optical VPN creation, network slicing, and optical interface management.
  • Hypervisors for VONs: OpenSlice dynamically provisions end-to-end paths and slices elastic-optical spectrum through extended OpenFlow messages and specialized optical components.Its evaluation reports that OpenFlow outperforms GMPLS for paths with more than three hops.
  • Application-Layer SDON Services: Access-network virtualization studies primarily exploit optical physical-layer and MAC-layer properties, including OFDMA subcarriers and polling-based MAC protocols.
  • Application-Layer SDON Services: Time-aware SDN applications dynamically balance data-center load using traffic correlations and report improvements in network blocking probability, global blocking probability, and spectrum consumption ratio.The study did not consider short-time-scale traffic bursts.
  • Application-Layer SDON Services: Bandwidth provisioning that maintains each application’s minimum “sweet point” bandwidth can significantly improve QoE while other applications receive their required bandwidth.

3) Virtual Topology Reconfigurations:

SDON application studies reconfigure virtual topologies, reserve bandwidth, optimize routing, manage access, and cache video content to address QoS and service-delivery requirements.

  • Virtual Topology Reconfigurations: SDN-controlled virtual-topology reconfiguration uses localized traffic and optical-resource information to improve network utilization and throughput.The architecture separates the SDN controller’s abstract network view from a WDM controller that configures ROADMs and lightpaths.
  • QoS Routing: Integrated access, metro, and core routing uses OpenFlow, I2RS, and PCE interactions, with experiments validating optimization based on current traffic and previous load.
  • QoS Routing: QoS-aware MPTCP bandwidth reservation performs path selection followed by OBS wavelength reservation, assigning higher priorities to latency-sensitive flows.
  • Video Applications: Video caching in software-defined PONs can service up to 30% additional videos while reducing service response delay to 50%.The mechanism jointly provisions bandwidth and pushes frequently requested content to base-station caches, including via multicast.
  • Access Control and Security: FlowNAC uses SDN-controller-installed OpenFlow forwarding rules to regulate service access using identifiers such as MAC addresses and IP source and destination addresses.

2) Lightpath Hopping Security:

The survey covers SDON security, energy-saving control, and failure recovery, including lightpath hopping, link shutdown, backup reprovisioning, and multidomain restoration.

  • Lightpath Hopping Security: Optical lightpath hopping changes a flow’s channel sequence, while guard periods separate transmissions by distinct flows sharing a channel.
  • Energy Efficiency: SDN’s separated control and data planes plus global network perspective support applications that monitor and reduce optical-network energy consumption.
  • Energy-Saving Routing: Link switch-off techniques must closely monitor QoS during link shutdown and flow rerouting.
  • Energy-Saving Routing: Data-center simulations report energy savings above 80% at low utilization and below 40% at high utilization when switches and hosts can be turned off.
  • Failure Recovery: Backup reprovisioning computes primary and link-disjoint backup paths, stores backups as logical links, and uses degraded capacity to accommodate more requests.Emulations indicate improved network utilization while provisioning restoration paths after failures.
  • Failure Recovery: ABNO supports multidomain recovery by pre-computing backup connections and delegating end-to-end configurations to lower-layer domain SDN controllers.Recovery requirements on large-scale networks are described as challenging at tens-of-milliseconds timescales.

4) Hierarchical Survivability:

SDON survivability and orchestration use centralized or hierarchical control to coordinate multilayer resources, recovery, and application requirements. The surveyed work highlights stronger recovery coordination but also limited study coverage and important unresolved QoS, security, and energy-efficiency concerns.

  • Hierarchical Survivability: Hierarchical controllers coordinate optical and IP resources for alternate-route recovery, but may incur long restoration times.Cross-layer restoration can recover from multipoint and concurrent failures.
  • QoS Applications: SDON QoS studies monitor traffic or network states through OpenFlow and centralized controllers to react to changing demands and conditions.Existing studies mainly address traffic and network management mechanisms.
  • QoS Applications: Future QoS studies should cover more application traffic types, including VoIP, whose low bit-rate traffic still requires low end-to-end latency.Video streaming has received attention, while other prominent traffic types remain less explored.
  • Open Challenges: Very few SDON studies address security, access control, or energy efficiency, leaving both as important future research directions.The survey identifies broad security and privacy challenges alongside limited attention to energy efficiency.
  • Multilayer Orchestration: Multilayer networking may be vertical within one domain or horizontal across domains, using SDN to coordinate different technologies and layers.Examples include IP, MPLS, and WDM, with SDN orchestrators coordinating layered technology regions.

2) Application-specific Orchestration:

Application-specific SDON orchestration integrates multilayer and multidomain control to improve recovery, resource utilization, virtualization, and service provisioning. Across domains, studies balance centralized coordination against distributed autonomy while handling heterogeneous technologies and operational complexity.

  • Failure Recovery: Multilayer SDN integration targets failure recovery by improving resource utilization while maintaining high protection levels and controlling network costs.The approach combines WDM and IP layers and is informed by a five-year backbone failure observation study.
  • Resource Utilization: Centralized SDN control integrates IP and WDM topology management with spectrum allocation to improve resource utilization and reduce transmission latencies.The orchestrator supports dynamic and simultaneous network-control operations through service abstraction.
  • Virtual Optical Networks: Virtual optical networks virtualize physical resources within domains and are hierarchically managed by an SDN network orchestrator.The framework integrates multiple transport technologies, including IP and GMPLS, with network virtualization controllers.
  • Multidomain Orchestration: Multidomain orchestration seeks resource-efficient end-to-end provisioning across distributed domains with heterogeneous layer technologies.Large deployments make cross-domain, cross-layer service provisioning highly challenging.
  • Hierarchical Multidomain Control: Hierarchical multidomain control uses parent and domain controllers to abstract physical resources across multiple vendors and optical transport technologies.Domain controllers virtualize physical-layer resources for coordination by the parent controller.
  • Inter-Domain Protocol: Flat multidomain control uses an Inter-Domain Protocol to coordinate autonomous domain controllers during cross-domain lightpath setup.The associated Routing and Spectrum Allocation algorithm supports end-to-end provisioning across multiple domains, with experimental validation across domains in the USA and China.
  • Multidomain Virtualization: Multidomain network hypervisors create customer-specific virtual network slices managed by customer SDN controllers over domain-specific physical controllers.The hypervisor operates above the network orchestrator and physical SDN controllers.
  • Control Trade-offs: Multidomain orchestration coordinates complex management decisions, trading centralized hierarchical decision making against distributed control.The surveyed studies report greater coordination capability alongside operational complexity from additional network entities.

VIII. OPEN CHALLENGES AND FUTURE SDON RESEARCH DIRECTIONS

Future SDON research must address cross-cutting challenges in integrating heterogeneous optical infrastructure, simplifying centralized management, supporting applications, and ensuring reliability, security, scalability, and standardization. The survey also identifies the need for solutions that enable SDON deployment in operational networks.

  • Cross-cutting integration: Heterogeneous devices and vendors make complete optical networking solutions costly to integrate, maintain, and manage.Some hardware components lack complete software stacks, requiring providers to maintain software development teams for integration.
  • Simplicity and efficiency: SDON management should combine central control with simple, efficient automation across network architectures and protocol layers.The survey notes that overly complex solutions risk high expenditures and are generally unlikely to be deployed.
  • NorthBound Interface: NorthBound Interface research must support application policies, dynamic service deployment, and communication across multiple controllers and optical-layer interfaces.Applications express service policies through the NBI, while future APIs should make optical components amenable to efficient controller communication.
  • Reliability, security, and privacy: Centralized SDN control creates reliability, security, and privacy challenges because control failures or infringements can cause extensive disruption.Future work should refine automated diagnostics, monitoring, and reconfiguration for rapid failure recovery.
  • Scalability and standardization: Scalable SDON deployment requires economical overlays for existing non-SDN infrastructure and standardization that supports multivendor interoperability.Open-source software and common architectures or protocol configurations are proposed as groundwork for standardization.

F. Multilayer Networking

SDON multilayer networking must coordinate optical-specific parameters with higher-layer control across single and multiple domains, while supporting heterogeneous optical-wireless networks and application-specific QoS. The survey identifies orchestration, access control, scalability, and comprehensive evaluation as continuing research needs.

  • Vertical multilayer networking: Optical multilayer networking differs from conventional IP SDN because it must provision parameters such as wavelengths and signal strengths.Optical orchestrators can provide a single interface between the SDN controller and optical devices.
  • Control interfaces: South-bound renderers let controllers communicate through multiple SBI protocols, including OpenFlow, SNMP, and NETCONF, enabling hybrid networks.This supports operation with both OpenFlow-enabled and traditional non-OpenFlow switches.
  • Multidomain networking: Multidomain SDON requires mechanisms for access control, authentication, authorization, accounting, and routing across organizational boundaries.Intent APIs are identified as a possible way to translate application intents into cross-domain service configurations.
  • Optical-wireless networking: FiWi networks need specialized multilayer strategies because wireless links have variable channel quality, higher losses, lower bit rates, and mobile end nodes.The survey connects this need with advances in femtocells, heterogeneous cellular structures, machine-to-machine communication, and energy savings.
  • QoS and energy efficiency: SDON should adapt QoS and energy use to application requirements, balancing high-rate video traffic against the low-latency needs of VoIP and videoconferencing.Broadcasting can require data rates from 3–48 Gb/s and also requires multicast-group management.
  • Performance evaluation: Evaluation methodologies should measure data-plane, control-plane, interface, virtualization, and orchestration performance against conventional networks.Benchmarking is intended to quantify the flexibility-related tradeoffs and costs of SDN and virtualization.
Loading 1511.04376v3…