Optical Networks and Switching

Introduction and Context

Optical communication has evolved from point-to-point transmission links into programmable, multi-layer networks that interconnect access users, metropolitan aggregation sites, data centers, cloud platforms, and national or international backbones. The central engineering problem is no longer merely transmitting a high-rate signal over fiber. It is the coordinated design of capacity, topology, switching, routing, protection, latency, energy consumption, and service quality.

Modern optical networks combine passive optical networks (PONs) in the access segment, packet and optical transport technologies in metro networks, reconfigurable optical add-drop multiplexers (ROADMs) in the backbone, and high-capacity optical switching fabrics in data centers. Optical transport networking (OTN) provides digital framing, multiplexing, monitoring, and fault management, while wavelength-division multiplexing (WDM) and elastic optical networking improve fiber utilization.

This chapter develops the architecture and engineering principles required to design these networks. Particular attention is given to routing and wavelength assignment, protection and restoration, quality of service, and the trade-offs between circuit, wavelength, spectrum, and packet switching.

Learning Objectives

After studying this chapter, the reader should be able to:

1. Explain the roles of access, metro, backbone, and data-center optical networks.

2. Compare PON, OTN, ROADM, fixed-grid WDM, and elastic optical networking.

3. Formulate routing and wavelength or spectrum assignment problems.

4. Distinguish protection from restoration and select suitable survivability schemes.

5. Evaluate optical switching technologies and their application areas.

6. Design a representative multi-layer network subject to capacity, reach, latency, and availability constraints.

7. Assess operational, regulatory, security, energy, and economic implications.

1. Network Segments and Design Requirements

An optical network is commonly divided into four segments.

Access networks connect subscribers, homes, enterprises, and radio sites to an aggregation point. They require low cost per user, passive outside-plant equipment, efficient sharing, and simple maintenance. PONs are the dominant optical access architecture.

Metro networks aggregate traffic from access systems, data centers, mobile sites, and business customers. They typically have shorter distances, many traffic sources, and rapidly changing demand. Metro networks often combine Ethernet, packet switching, coherent optics, and ROADMs.

Backbone or core networks connect cities and major data centers over tens to thousands of kilometres. They emphasize very high capacity, optical reach, restoration, low blocking probability, and operational visibility.

Data-center networks interconnect servers, storage, accelerators, and external sites. Their defining requirements are high port density, low latency, fast reconfiguration, short reach, and low energy per bit.

A design must satisfy several constraints simultaneously:

• Capacity: offered traffic and growth must be supported without unacceptable blocking.
• Reach: the optical signal must meet power, dispersion, nonlinear, and OSNR requirements.
• Latency: propagation, switching, queuing, and processing delays must meet application requirements.
• Availability: failures of fiber, equipment, power, or software must be tolerated.
• Scalability: the architecture should accommodate additional users, wavelengths, and sites.
• Operational simplicity: provisioning, monitoring, and fault isolation should be practical.
• Economic efficiency: equipment, fiber, energy, spectrum, and maintenance costs must be balanced.

2. Passive Optical Networks

2.1 Architecture and operation

A PON uses an optical line terminal (OLT) at the central office, an optical distribution network (ODN), and optical network units or terminals (ONU/ONT) at customer premises. The ODN contains feeder fiber, splitters, distribution fiber, and drop fiber. Splitters are passive, so they require no local power and can serve many users from one OLT port.

Downstream transmission is generally broadcast from the OLT to ONUs, with logical identification and encryption preventing unauthorized reception. Upstream transmission is shared. The OLT therefore allocates transmission windows using dynamic bandwidth allocation (DBA). Time-division multiplexing avoids collisions, while ranging compensates for different propagation delays from ONUs.

PON performance is governed by split ratio, fiber loss, connector and splice loss, transmitter power, receiver sensitivity, and the optical power budget. A simplified power-budget condition is

P_T - P_R \ge L_F + L_S + L_C + M,

where \(P_T\) is minimum transmitter power in dBm, \(P_R\) is receiver sensitivity in dBm, \(L_F\) is fiber attenuation, \(L_S\) is splitter loss, \(L_C\) is connector and splice loss, and \(M\) is an engineering margin. All terms are expressed in dB or dBm consistently. The relation applies as a first-order link-budget test; dispersion, nonlinear effects, wavelength dependence, and burst-mode receiver performance require additional analysis.

For fiber length \(L\) and attenuation coefficient \(\alpha\) in dB/km, fiber loss is

\[L_F = \alpha L.\]

A practical splitter loss is greater than the ideal splitting loss because of excess loss. For an ideal \(N\)-way splitter,

\[L_{split,ideal}=10\log_{10}(N)\ \text{dB}.\]

For example, an ideal 1:32 splitter has a splitting loss of approximately 15.05 dB. Real devices add excess loss.

2.2 PON evolution

Broadband PON systems have progressed from GPON and EPON toward 10G-class systems such as XG-PON and XGS-PON, and toward higher-rate generations including 25G, 50G, and future coexistence schemes. Coexistence is achieved by assigning different downstream and upstream wavelength bands, allowing multiple generations to share portions of the outside plant.

PONs are attractive because their passive field infrastructure reduces power consumption and maintenance. Their limitations include shared bandwidth, split-loss constraints, upstream scheduling delay, limited physical-layer flexibility, and a common dependency on feeder and splitter infrastructure. Protection can be added through dual homing, redundant feeder fibers, or protected OLT configurations, but these increase cost.

3. Optical Transport Networking and Metro Architecture

OTN is a digitally managed transport framework standardized by ITU-T recommendations. It provides client-independent containers, forward error correction (FEC), hierarchical multiplexing, performance monitoring, tandem connection monitoring, and fault-management functions. Client signals such as Ethernet, storage traffic, and video can be mapped into OTN containers and transported over optical channels.

OTN separates the digital layer, where framing, multiplexing, and monitoring occur, from the optical layer, where wavelengths, amplifiers, ROADMs, and fibers operate. This separation enables operators to supervise services even when the underlying optical path changes.

A typical OTN hierarchy contains lower-order and higher-order containers. The exact rates and mappings depend on the standard and implementation. In practice, a 100-Gbit/s client may be carried in an appropriate OTN digital container and then assigned to an optical channel. FEC adds overhead but improves achievable reach by allowing correction of transmission errors.

Metro networks commonly use a ring, partial mesh, or hub-and-spoke topology. Rings simplify protection and operations, whereas meshes provide better resource sharing and alternate paths. Packet-optical convergence is common: Ethernet switches perform statistical multiplexing and service functions, while coherent optical transponders provide long-reach transmission.

4. WDM, ROADMs, and Optical Cross-Connects

WDM transmits multiple optical channels on one fiber by assigning each channel a distinct wavelength or frequency slot. Dense WDM uses closely spaced channels and optical amplifiers to achieve high fiber capacity.

A ROADM can add, drop, or pass selected wavelengths without converting every channel to electronics. Its principal functions include:

• remotely configurable wavelength provisioning;
• express passage of channels through a node;
• wavelength add/drop;
• power equalization and monitoring;
• rapid service reconfiguration;
• support for mesh protection and restoration.

Colorless ROADMs allow a port to handle different wavelengths. Directionless ROADMs permit a wavelength to be routed toward different fiber directions. Contentionless designs reduce conflicts when multiple channels use the same wavelength at a node. These features improve flexibility but increase optical complexity, power consumption, and control requirements.

An optical cross-connect (OXC) establishes connections among fibers, wavelengths, or spectrum slots. Switching may be implemented using micro-electromechanical systems (MEMS), wavelength-selective switches (WSSs), liquid-crystal technologies, semiconductor optical amplifiers, or integrated photonic circuits. Large optical switching fabrics must manage insertion loss, crosstalk, wavelength selectivity, switching time, and control stability.

5. Fixed-Grid and Elastic Optical Networking

Traditional WDM uses a fixed frequency grid, historically with channel spacings such as 50 GHz. A channel is assigned a fixed spectral width even when its modulation format and traffic demand do not require the entire slot.

Elastic optical networking (EON) divides the spectrum into fine-grained frequency slots. A connection is assigned a contiguous group of slots sufficient for its bit rate, modulation format, forward-error-correction overhead, and guard band. EON improves spectral efficiency but introduces the spectrum-continuity and spectrum-contiguity constraints.

If a connection uses path \(p\), frequency-slot set \(S\), and modulation format \(m\), feasibility requires:

1. every link on \(p\) has the slots in \(S\) available;

2. the slots are contiguous;

3. the same slot indices are available on all links unless spectrum conversion is provided;

4. the selected modulation format supports the path length and optical impairment budget;

5. guard bands separate adjacent channels when required.

The required number of slots can be represented generally as

\[n_s = \left\lceil\frac{B_{req}}{B_{slot}}\right\rceil + n_g,\]

where \(B_{req}\) is the required occupied bandwidth, \(B_{slot}\) is the slot width, and \(n_g\) represents guard-slot requirements. The value of \(B_{req}\) depends on symbol rate, modulation, FEC, shaping, and implementation; it must not be inferred from nominal data rate alone.

EON introduces the distance-adaptive principle: shorter paths may use higher-order modulation and fewer slots, while longer paths use more robust modulation and more slots. This creates a trade-off between spectral efficiency and reach.

6. Routing and Wavelength Assignment

In a wavelength-routed network, a connection request must be assigned a path and wavelength. The classical routing and wavelength assignment (RWA) problem has two constraints:

• wavelength continuity: the same wavelength is used throughout the path when no wavelength converter exists;
• wavelength noncollision: two connections sharing a link cannot use the same wavelength.

A common practical strategy is fixed-alternate routing followed by first-fit wavelength assignment. Candidate paths are ordered, and the first wavelength available on every link is selected. First-fit is computationally simple and often performs well, although it is not universally optimal.

For dynamic traffic, the objective may be to minimize blocking probability, maximize carried traffic, reduce power, or preserve resources for high-priority services. For static planning, the problem may be formulated as an integer optimization model with variables indicating whether a demand uses a particular path and wavelength. Exact formulations become computationally difficult as network size increases, so heuristics, decomposition, and metaheuristics are widely used.

In EON, RWA becomes routing, modulation, and spectrum assignment (RMSA). The algorithm must select a route, modulation format, and contiguous spectrum segment. It must also consider fragmentation: free spectrum may exist in total but be unusable because it is divided into small noncontiguous gaps.

Control can be centralized through a software-defined optical network controller or distributed through signaling and routing protocols. Centralized control improves global optimization, while distributed control can improve resilience and reduce dependence on a single controller.

7. Protection, Restoration, and Availability

Protection reserves resources in advance for a failure. On detecting a fault, traffic switches to a preplanned backup path. Dedicated 1+1 protection sends traffic simultaneously over working and protection paths, with the receiver selecting the better copy. Shared-mesh protection reserves backup capacity that may be shared among demands whose working paths are unlikely to fail simultaneously.

Restoration calculates or establishes an alternate path after a failure. It can use resources more efficiently than dedicated protection but generally has longer recovery time and greater control-plane dependence.

Protection architectures include:

• ring protection for simple metro deployments;
• link protection for rapid recovery from a single link failure;
• path protection for end-to-end service continuity;
• segment protection for large networks;
• multilayer protection using packet, OTN, and optical layers.

Independent protection actions at several layers can cause unnecessary capacity reservation or oscillatory switching. A coordinated multilayer strategy should determine which layer responds to a fault and suppress duplicate recovery actions.

For independent component availabilities, a simple series-path approximation is

\[A_{path}=\prod_{i=1}^{n} A_i,\]

where \(A_i\) is the availability of component \(i\). This assumes statistically independent failures and a series service model. Real networks may violate these assumptions because a single cable cut can affect multiple fibers, and common power or duct failures create correlated outages.

8. Optical Switching in Data Centers

Data-center switching occurs at several time scales. Electronic packet switches offer buffering, mature control, and fine-grained statistical multiplexing. Optical circuit switches provide high capacity and low data-plane energy for stable large flows, but circuit setup time and the absence of simple optical buffering make them less suitable for highly bursty traffic. Optical burst and packet switching attempt faster adaptation but require demanding contention-resolution, synchronization, and control mechanisms.

Hybrid electrical-optical architectures are therefore common. Electronics handle fine-grained and bursty traffic, while optical circuits carry predictable elephant flows, inter-rack aggregates, or accelerator-to-accelerator traffic. Emerging silicon photonics and co-packaged optics can reduce electrical interconnect distance and energy, although thermal management, manufacturing yield, packaging, and field replacement remain important issues.

9. Quality of Service and Performance Engineering

Optical networks support QoS through multiple mechanisms:

• traffic engineering and constrained routing;
• differentiated protection classes;
• bandwidth reservation;
• latency-aware path selection;
• packet prioritization at electronic layers;
• optical impairment-aware admission control;
• service-level monitoring.

Important metrics include blocking probability, throughput, latency, jitter, packet loss, restoration time, availability, OSNR margin, bit-error ratio, and energy per transported bit. Optical layers do not automatically provide packet-level QoS. A low-loss optical path may still experience congestion in transponders, switches, or queues.

A service request should therefore be mapped across layers. For example, a latency-sensitive financial or industrial service may require a short physical route, dedicated capacity, rapid protection, and strict monitoring. A backup service may tolerate longer restoration and use shared protection.

10. Illustrative Design Case: Regional Cloud and Broadband Network

Background

Consider an illustrative operator serving 20,000 broadband subscribers in three towns, two enterprise data centers, and one mobile-radio aggregation site. The towns are 35–60 km from a regional core hub. Demand is growing rapidly because of video, cloud access, and 5G transport.

Problem

The operator must provide economical access, connect the towns to both data centers, survive a single fiber cut, and support future 400-Gbit/s backbone channels. A simple point-to-point design would require excessive transponders and would be difficult to reconfigure.

Analysis

The access segment has many users and moderate individual traffic, making a shared PON appropriate. The metro segment has aggregation and protection requirements, favoring a packet-optical ring with ROADM-enabled interconnection. The core requires mesh routing and wavelength or spectrum assignment. Because future traffic includes long-distance high-capacity services, reach-adaptive coherent transmission is required.

Approach and solution

Each town is served by multiple PON OLT ports. Split ratios are selected using the optical power budget rather than subscriber count alone. The OLTs are dual-homed where the business case justifies protection. Traffic is aggregated into OTN containers at metro sites and carried over a two-fiber ROADM ring. The two data centers are connected through diverse fiber routes.

At the core, working paths are computed using shortest feasible routes, while backup paths are link-disjoint where fiber geography permits. High-priority services receive preplanned protection; lower-priority bulk traffic uses restoration. For 400-Gbit/s demands, the controller evaluates modulation format, path length, OSNR margin, and contiguous spectrum. A centralized controller maintains a topology and resource database, while local devices retain safe fallback behavior if controller connectivity is lost.

Outcome

The design reduces access equipment cost through passive splitting, enables remote wavelength provisioning in the metro, and avoids reserving dedicated protection capacity for every low-priority service. It also exposes a key engineering reality: logical diversity is not physical diversity. Two routes that appear different in the control plane may share a duct, bridge, or power facility.

Learning points

1. Network segmentation allows each technology to be matched to its traffic and distance characteristics.

2. Protection decisions must consider geography, not only graph topology.

3. Capacity planning must include growth and restoration traffic.

4. Optical flexibility is valuable only when operational processes and monitoring can use it.

5. Multilayer coordination prevents inefficient or conflicting recovery actions.

11. Challenges, Governance, and Contemporary Developments

Optical networks face several limitations. Fiber cuts remain a major source of outages. High-order modulation improves spectral efficiency but reduces reach and impairment margin. Flexible-grid systems can suffer spectrum fragmentation. ROADMs and coherent transponders increase programmability but also increase software and configuration complexity.

Automation and artificial intelligence are increasingly used for traffic prediction, anomaly detection, optical impairment estimation, failure prediction, and closed-loop restoration. AI recommendations should remain subject to engineering constraints, explainability requirements, and safe rollback. Training data may be biased toward normal operating conditions and fail during rare correlated failures.

Security concerns include unauthorized controller access, manipulation of optical configuration, denial of service through false alarms, compromised telemetry, and attacks on management networks. Strong authentication, role-based authorization, secure interfaces, configuration audit trails, segmentation, and independent fail-safe controls are essential.

Regulatory and governance issues include spectrum coordination, lawful access obligations, critical-infrastructure resilience, data protection, environmental reporting, and interoperability. Energy efficiency should be evaluated across the complete life cycle, including equipment manufacture, cooling, operation, replacement, and disposal.

Best practice is to maintain an accurate physical inventory, document fiber-route diversity, validate optical margins under worst-case conditions, use staged deployment, monitor performance continuously, test protection mechanisms, and conduct failure drills. Design decisions should be based on measured traffic and failure data rather than nominal equipment capacity alone.

Chapter Summary

Optical networks are multilayer systems spanning PON access, packet-optical metro aggregation, OTN transport, ROADM-based backbone networks, elastic optical systems, and data-center switching fabrics. PONs reduce access cost through passive sharing and DBA, while OTN supplies digital framing, monitoring, multiplexing, and FEC. ROADMs and OXCs enable wavelength-level reconfiguration, and elastic optical networking improves spectral efficiency through flexible slots and reach-adaptive modulation.

Routing and wavelength assignment must satisfy path, continuity, collision, and capacity constraints; in elastic networks, modulation and spectrum contiguity are additional concerns. Protection reserves resources in advance, whereas restoration computes recovery after failure. Availability calculations require careful attention to independence and common-cause failures. QoS depends on coordinated traffic engineering across optical, OTN, packet, and application layers.

Practical network design requires balancing capacity, reach, latency, availability, energy, cost, operational complexity, and physical diversity. Automation and AI can improve prediction and control, but they must operate within secure, explainable, and fail-safe governance frameworks.

Key Terms

PON: A shared passive optical access network using an OLT, passive splitters, and ONUs.

OLT: Optical line terminal that controls and transmits to a PON.

ONU/ONT: Customer-side optical network unit or terminal.

Dynamic bandwidth allocation: Scheduling method that assigns shared upstream PON capacity.

OTN: Digitally managed optical transport framework with framing, multiplexing, monitoring, and FEC.

WDM: Transmission of multiple optical channels using different wavelengths or frequencies.

ROADM: Reconfigurable optical add-drop multiplexer.

OXC: Optical cross-connect that establishes optical paths among fibers or channels.

WSS: Wavelength-selective switch used to route selected optical channels.

Optical reach: Maximum feasible transmission distance under specified impairment and margin conditions.

RWA: Routing and wavelength assignment.

RMSA: Routing, modulation, and spectrum assignment.

Wavelength continuity: Requirement to retain one wavelength along a path without conversion.

Elastic optical network: Network using flexible frequency slots rather than only fixed channel grids.

Protection: Preplanned recovery using reserved backup resources.

Restoration: Recovery established after a failure occurs.

ROADM colorless operation: Ability of a port to handle different wavelengths.

Quality of service: Performance differentiation according to service requirements.

Spectrum fragmentation: Condition in which free spectrum exists but is divided into unusable gaps.

Physical diversity: Independence of actual fiber routes, ducts, power systems, and facilities.

Review Questions

Short Answer Questions

1. What are the principal components of a PON?

2. State the difference between protection and restoration.

3. What is the wavelength-continuity constraint?

4. Why are ROADMs important in mesh optical networks?

5. Define spectrum fragmentation.

Long Answer Questions

1. Explain PON operation, including downstream broadcast, upstream DBA, ranging, and optical power budgeting.

2. Compare fixed-grid WDM and elastic optical networking in terms of capacity, control, and limitations.

3. Describe OTN architecture and explain how it complements the optical layer.

4. Discuss protection architectures and the problem of multilayer recovery coordination.

5. Explain how optical switching technologies differ in access, metro, backbone, and data-center applications.

Analytical / Application Questions

1. Design a protection and routing strategy for a three-site metro network carrying premium and best-effort services. Justify the use of protection or restoration for each class.

2. A proposed EON route has sufficient total free spectrum but no contiguous block across all links. Explain why the demand is blocked and propose two remedies.

3. Develop a preliminary architecture for a regional operator serving homes, mobile sites, and two data centers. Select technologies for each segment and identify the principal capacity, reach, security, and availability risks.

Multiple Choice Questions with Answers

Question 1

Which function is primarily associated with upstream operation in a PON?

A. Dynamic bandwidth allocation

B. Wavelength conversion at every ONU

C. Optical amplification in each splitter

D. Dedicated fiber for every subscriber

Correct Answer: A. Dynamic bandwidth allocation

Question 2

What is the ideal splitting loss of a 1:32 optical splitter?

A. 3.01 dB

B. 10.00 dB

C. 15.05 dB

D. 32.00 dB

Correct Answer: C. 15.05 dB

Question 3

Which technology provides digital framing, multiplexing, monitoring, and forward error correction?

A. ROADM

B. OTN

C. WSS

D. PON splitter

Correct Answer: B. OTN

Question 4

In a network without wavelength conversion, wavelength continuity means that:

A. Every link must use a different wavelength

B. The same wavelength is retained along the complete optical path

C. All wavelengths are converted to a common electrical signal

D. Only one wavelength can exist in the network

Correct Answer: B. The same wavelength is retained along the complete optical path

Question 5

What is a principal advantage of a ROADM?

A. It eliminates all fiber attenuation

B. It permits remote reconfiguration of optical channels

C. It guarantees unlimited reach

D. It replaces every packet switch

Correct Answer: B. It permits remote reconfiguration of optical channels

Question 6

Which additional decision is required in RMSA compared with classical RWA?

A. Subscriber authentication

B. Modulation-format selection

C. Splitter installation

D. IP address assignment

Correct Answer: B. Modulation-format selection

Question 7

How does protection generally differ from restoration?

A. Protection reserves recovery resources in advance

B. Protection can only be used in PONs

C. Restoration always has zero switching delay

D. Restoration does not require an alternate route

Correct Answer: A. Protection reserves recovery resources in advance

Question 8

A path contains independent components with availabilities 0.99 and 0.98. Under the series-path approximation, its availability is:

A. 0.9702

B. 0.9850

C. 1.9700

D. 0.0198

Correct Answer: A. 0.9702

Question 9

Why can a connection be blocked in an elastic optical network despite adequate total free spectrum?

A. Free spectrum may not be contiguous across every link

B. Elastic networks cannot use coherent transmission

C. All spectrum must be assigned to one connection

D. ROADMs operate only at one wavelength

Correct Answer: A. Free spectrum may not be contiguous across every link

Question 10

Which data-center traffic is most suitable for an optical circuit in a hybrid electrical-optical architecture?

A. A stable, high-volume elephant flow

B. A randomly arriving single-packet flow

C. A management alarm of a few bytes

D. A highly bursty flow with no predictable duration

Correct Answer: A. A stable, high-volume elephant flow