Wavelength Division Multiplexing

A DWDM system connecting two cities is impressive engineering—but it's just a link. To build the global optical infrastructure connecting thousands of locations, we need to elevate from links to networks: interconnected systems where wavelengths can be routed, protected, and managed across complex topologies.

Optical networking transforms individual wavelength channels into a flexible, highly available transport fabric. Instead of static point-to-point connections, modern optical networks can:

• Route wavelengths through multiple intermediate nodes without electrical conversion
• Automatically reroute traffic in milliseconds when fibers are cut
• Add or remove capacity on specific routes without affecting others
• Balance load across topology to maximize utilization

This evolution—from opaque, electrically-switched networks to transparent, optically-switched networks—represents one of telecommunications' most profound architectural shifts. Understanding optical networks is essential for anyone building or managing modern communications infrastructure.

Optical networking has evolved through distinct generations, each increasing flexibility and reducing operational complexity.

Generation 1: Point-to-Point WDM (1990s)

Early WDM systems were simple terminal-to-terminal links:

• Fixed wavelengths assigned at installation
• Any routing or switching done electrically at terminal sites
• High capacity but zero flexibility
• Every wavelength traveled the entire link—no intermediate access

Generation 2: Optical Add-Drop (Late 1990s-2000s)

Fixed Optical Add-Drop Multiplexers (FOADM) introduced intermediate access:

• Specific wavelengths could exit or enter at intermediate sites
• Remaining wavelengths passed through transparently
• Wavelength assignments still fixed at installation
• Required manual fiber patching to change configuration

Generation 3: Reconfigurable OADMs (2000s-2010s)

Reconfigurable Optical Add-Drop Multiplexers (ROADMs) brought software-controlled wavelength routing:

• Any wavelength can be dropped, added, or passed through
• Configuration changes via network management software
• No on-site technician required for wavelength changes
• Revolutionary for operational efficiency

Generation 4: Colorless-Directionless-Contentionless ROADMs (2010s-Present)

CDC (Colorless, Directionless, Contentionless) ROADMs removed final limitations:

• Colorless: Any wavelength at any add/drop port (vs. fixed-wavelength ports)
• Directionless: Any wavelength can be routed to any direction (vs. fixed line-direction associations)
• Contentionless: Two signals on the same wavelength can be handled simultaneously (same wavelength from different directions routed to different destinations)

Generation 5: Autonomous/Disaggregated Networks (Emerging)

Current evolution splits hardware and software:

• Disaggregated optical: Line system and transponders from different vendors
• Open ROADMs: Standardized interfaces via OpenROADM MSA
• SDN control: Centralized software controls distributed optical elements
• Intent-based networking: Operators specify outcomes, not configurations

ROADMs are the fundamental building blocks of modern optical networks. Understanding their internal architecture reveals how flexible wavelength routing is achieved.

Core ROADM Functions:

1. Express (pass-through): Wavelengths transit the node without processing
2. Drop: Wavelengths exit the network for local termination
3. Add: Locally-sourced wavelengths enter the network
4. Route: Wavelengths are directed between input and output degrees

ROADM Degree and Architecture:

A degree represents a fiber direction—a pair of fibers (one transmit, one receive) connecting to an adjacent node. ROADMs are characterized by their degree count:

• 2-degree: Linear (pass-through) topology, two fiber directions
• 4-degree: Common ring interconnect, four fiber directions
• 8-degree: Large junction nodes connecting many routes
• 16+ degree: Mega-nodes at major data center locations

Wavelength Selective Switches (WSS):

The WSS is the core technology enabling ROADMs. A WSS can independently route each wavelength channel to different output ports:

• 1×N WSS: One input, N outputs (or reverse: N inputs, one output)
• Switching granularity: Per-wavelength or per-slot (flex grid)
• Attenuation control: Per-channel power equalization
• Technology: MEMS mirrors, LCoS (Liquid Crystal on Silicon), or silicon photonics

Example: 1×20 WSS at 96 channels

• Each of 96 wavelengths independently directed to any of 20 output ports
• Equivalent to 96 separate switches, implemented in one device
• Switching time: milliseconds (MEMS) to microseconds (LCoS)

CDC Implementation:

Colorless: Transponder ports connect through tunable filters or WSS—any transponder can emit any wavelength.

Directionless: Add/drop structure fans out to all degrees through a WSS—dropped signals from any direction reach the same transponder pool.

Contentionless: Multiple WSS planes separate same-wavelength signals from different directions—if λ₁ arrives from Degree 1 and Degree 3, they can be routed to different destinations or dropped independently.

Twin (CDC-T): Adds ability to route two instances of the same wavelength to the same output fiber (for protection), using additional WSS planes.

In optical networks, wavelengths are the fundamental resource being routed. The problem of assigning wavelengths to connections is known as the Routing and Wavelength Assignment (RWA) problem—one of optical networking's central challenges.

The RWA Problem:

Given a set of connection requests (source-destination pairs), RWA finds:

1. Route: Physical path through the network for each connection
2. Wavelength: Which lambda carries each connection

Constraints:

• Wavelength continuity: Without conversion, a signal must use the same wavelength end-to-end
• Wavelength clash: Two lightpaths on the same fiber cannot use the same wavelength
• Physical impairments: Path must have acceptable OSNR, dispersion, nonlinearity

RWA Complexity:

RWA is NP-complete for general networks. Heuristics and approximations handle practical cases:

Static RWA: Pre-compute routes for known traffic matrix (network planning) Dynamic RWA: Compute routes as requests arrive (real-time provisioning)

Impairment-Aware RWA:

Modern optical networks incorporate physical layer constraints into path computation:

Quality of Transmission (QoT) Estimation:

• Calculate expected OSNR at receiver based on: span losses, amplifier noise figures, nonlinear distortion, ROADM losses, coherent transceiver capabilities
• Only admit paths that exceed minimum OSNR threshold with margin

Margin Management:

• New connections consume network margin
• Tight margin = high utilization but fragile to degradation
• Generous margin = lower utilization but robust

Spectrum Fragmentation (Flex Grid):

With flexible grid systems, RWA becomes Routing and Spectrum Assignment (RSA):

• Allocate contiguous spectrum slots, not just single wavelengths
• Wide channels (400G, 800G) need 3-6 contiguous 12.5 GHz slots
• Over time, spectrum becomes fragmented—wide channels may not fit even with total capacity available
• Defragmentation periodically reorganizes spectrum (analogous to disk defragmentation)

Physical topology—how nodes and fibers are interconnected—fundamentally affects network capacity, resilience, and operational flexibility.

Ring Topology:

Historically dominant in SONET/SDH networks, rings remain common for metro optical:

Advantages:

• Simple protection: traffic flows both directions, automatic failover
• Minimal fiber needed (N fibers for N nodes)
• Easy to understand and manage

Disadvantages:

• Inefficient for mesh traffic patterns (signals route the "long way" around)
• Adding nodes disrupts ring
• Capacity limited by weakest link

Mesh Topology:

Modern long-haul and even metro networks increasingly use mesh:

Advantages:

• Shorter paths between nodes (direct vs. around ring)
• Multiple protection paths (higher resilience)
• Better scalability—add nodes with point-to-point connections
• Efficient spectrum utilization

Disadvantages:

• More fiber required (potentially N×(N-1)/2 links for full mesh)
• Complex routing decisions
• Harder to visualize/troubleshoot

Hub-and-Spoke:

Data center and enterprise networks often use hub-and-spoke:

Advantages:

• Simple connectivity model
• Centralizes expensive equipment at hub
• Easy provisioning (all traffic through hub)

Disadvantages:

• Hub is single point of failure
• Traffic between spokes must transit hub (inefficient)
• Hub becomes capacity bottleneck

Hierarchical Designs:

Large networks combine multiple topologies hierarchically:

Core Layer: Fully meshed high-capacity backbone connecting major metros Metro Core: Meshed rings within metropolitan areas
Metro Access: Traditional rings or hub-spoke connecting cell sites, enterprises On-Net Buildings: Hub-spoke within building, connecting to metro access

Multi-Layer Integration:

Optical transport integrates with higher layers:

• IP/MPLS: Routers use optical wavelengths as virtual links
• OTN (Optical Transport Network): Adds multiplexing and management within optical layer
• SDN Integration: Unified controller views both IP and optical resources

Disaggregated Architecture:

Traditional networks were vertically integrated (single vendor, proprietary). Modern disaggregation separates:

• Optical line system: ROADMs, amplifiers, fiber monitoring (one vendor)
• Transponders: Client-to-wavelength conversion (another vendor)
• Control plane: Network management software (potentially open-source)

This approach drives competition and flexibility but requires robust standardization.

Fiber cuts happen—backhoes, ships, earthquakes, vandalism. Optical networks must recover in milliseconds to meet service-level agreements demanding 99.999% (five nines) availability.

Protection vs. Restoration:

Protection: Pre-provisioned backup path, switchover is automatic and fast Restoration: Backup path computed and provisioned after failure, slower but more flexible

Protection Architectures:

1+1 Protection (Highest Availability):

Traffic is simultaneously transmitted on both working and protection paths:

• Receiver selects better signal
• Switchover is instantaneous—receiving side just starts using other input
• No signaling required between ends
• Uses 2× capacity: one fiber carrying copy, waiting for failure
• Used for critical services: financial trading, emergency services

1:1 Protection:

Backup path remains idle until needed:

• Can carry low-priority traffic during normal operation
• Signaling required to activate protection on failure
• Still 50% efficiency but better utilization than 1+1

Shared Mesh Protection:

Multiple working paths share common protection capacity:

• Only sufficient spare capacity to handle one failure at a time
• After failure recovery, backup re-released for sharing
• Complex planning: ensure any single failure can be protected
• Achieves 60-75% network utilization vs. 50% for dedicated

Restoration:

Compute and provision backup path after failure:

• No pre-reservation: all capacity available for working traffic
• Slowest recovery (seconds to minutes vs. milliseconds)
• May fail if no feasible path exists post-failure
• Often combined: fast protection for priority traffic, restoration for best-effort

Just as routers run protocols to exchange reachability and compute paths, optical networks have their own control plane for automating wavelength provisioning, protection, and monitoring.

Control Plane Functions:

1. Neighbor Discovery: Identify adjacent optical nodes and link properties
2. Resource Advertisement: Broadcast available wavelengths, link capacities, impairments
3. Path Computation: RWA algorithms determining routes and wavelength assignments
4. Connection Signaling: Set up and tear down lightpaths via ROADM configuration
5. Protection Coordination: Fast switchover signaling between nodes
6. Performance Monitoring: Track quality metrics for all active paths

GMPLS (Generalized Multi-Protocol Label Switching):

GMPLS extends MPLS signaling concepts to optical networks:

• RSVP-TE Extensions: Signal lightpath setup with wavelength labels
• OSPF-TE Extensions: Advertise optical link properties (wavelengths, impairments)
• Link Management Protocol (LMP): Manage control channel separate from data plane

Labels in Optical Context:

• Router MPLS: label = numeric identifier in packet header
• Optical GMPLS: "label" = wavelength, time slot, or fiber port

Software-Defined Optical Networking:

SDN (Software-Defined Networking) principles are increasingly applied to optical:

OpenROADM:

• Multi-source agreement defining YANG data models for ROADMs
• Standard APIs for transponders, amplifiers, optical monitoring
• Enables multi-vendor interoperability at optical layer

TAPI (Transport API):

• ONF standard for abstracting transport networks
• Defines topology, connectivity, path computation services
• Allows higher-layer applications to request optical resources

OpenConfig:

• Vendor-neutral YANG models for network devices
• Optical extensions cover transponders, amplifiers, OTN
• Enables streaming telemetry for real-time monitoring

Hierarchical Control:

Large networks often use layered controllers:

• Domain Controller: Manages ROADMs within a metro or region
• Super Controller: Coordinates across domains for end-to-end paths
• Orchestrator: Integrates optical with IP and service layers

Automation Use Cases:

• Zero-Touch Provisioning: New services established without manual intervention
• Automatic Restoration: Failure detection triggers pre-computed backup activation
• Load Balancing: Traffic redistributed based on real-time utilization
• Proactive Maintenance: Degrade trends trigger preemptive rerouting
• What-If Simulation: Test impact of changes before implementation

Optical networks serve diverse applications, each with distinct requirements driving different architectural choices.

Long-Haul/Backbone Networks:

Connect major metropolitan areas and countries:

• Distances: 1,000-10,000+ km
• Capacity: 10-100+ Tbps per fiber pair
• Technology: Coherent DWDM, C+L band, shared mesh protection
• Topology: Mesh connecting major PoPs
• Operators: National carriers, hyperscalers

Metro Core Networks:

Interconnect sites within metropolitan areas (10-200 km):

• Purpose: Aggregate access traffic, connect data centers, mobile backhaul hubs
• Capacity: 1-20 Tbps per fiber pair
• Technology: Coherent or direct-detect DWDM depending on distance
• Topology: Rings evolving to mesh

Data Center Interconnect (DCI):

Ultra-high-capacity links between data centers:

• Distances: Campus (< 2 km) to continental (2,000+ km)
• Capacity: Largest per-link capacity anywhere—multiple Tbps
• Requirements: Lowest possible latency, maximum bandwidth
• Technology: Leading-edge coherent (400G-800G first deployed here)

Submarine Cable Networks:

The most challenging optical environment:

• Distance: Trans-ocean routes up to 15,000 km
• Amplifier spacing: Every 50-80 km (limited by electrical power delivery over cable)
• Lifetime: 25-year design life with no mid-life upgrade to fiber
• Repeaters: Hermetically sealed, pressure-tolerant, powered via cable
• Capacity upgrades: Only terminal equipment can be upgraded over cable lifetime

Consortium vs. Private Cables:

• Traditional: Built by consortium of carriers sharing capacity
• Modern: Hyperscalers (Google, Meta, Microsoft) build private cables for exclusive use
• Economic shift: Content providers now own significant submarine capacity

5G Mobile Networks:

Wireless networks are increasingly optical-dependent:

• Fronthaul: Connects Radio Units to Distributed Units (strict latency: <100μs)
• Midhaul: Connects DU to Centralized Units (~1ms latency)
• Backhaul: Connects CU to core network (5-10ms acceptable)

Massive MIMO and higher frequencies drive 10-100× more fronthaul capacity than 4G, accelerating DWDM deployment into mobile transport.

Optical networks transform individual DWDM links into flexible, resilient infrastructure capable of supporting global connectivity. Mastering these concepts is essential for anyone working with telecommunications infrastructure.

Looking Ahead:

The final page of this module examines long-haul transmission—the engineering challenges of pushing optical signals across continental and transoceanic distances. You'll explore specific impairments that accumulate over thousands of kilometers, learn how submarine cable systems achieve 25-year reliability, and understand the technology roadmap for next-generation ultra-long-haul optical systems.