Communication Network Foundations and Architectures

Introduction

A communication network is an interconnected system that transfers information among computational, sensing, and control devices. Modern networks range from short-range personal networks to carrier-scale optical backbones and highly distributed cloud infrastructures. Their engineering involves more than selecting transmission media: designers must define service requirements, divide functions among protocols, allocate addresses and resources, control congestion, and ensure interoperability among independently manufactured systems.

This chapter establishes the architectural foundation for advanced communication-network study. It examines circuit, packet, and message switching; layered reference models; encapsulation and addressing; multiplexing; topologies; service models; standards organizations; and end-to-end design principles. Particular attention is given to queueing, delay, quality of service (QoS), and performance analysis because a network that is logically correct but unable to meet delay, loss, or throughput requirements is not an adequate engineering solution.

Learning Objectives

By the end of this chapter, a student should be able to:

1. Compare circuit, message, and packet switching using resource, delay, and traffic criteria.

2. Explain OSI and TCP/IP architectures and relate their layers to practical protocols.

3. Describe protocol layering, encapsulation, demultiplexing, and addressing.

4. Analyze multiplexing methods and common network topologies.

5. Distinguish network service models and evaluate their QoS implications.

6. Apply basic propagation, transmission, processing, queueing, and end-to-end delay models.

7. Use simple queueing relationships to interpret utilization, waiting time, and congestion.

8. Identify the roles of standards bodies and apply end-to-end design principles.

1. Communication Networks as Engineered Systems

A network provides communication services through coordinated functions. At minimum, a source produces data, an intermediate system forwards or switches it, and a destination consumes it. The network must represent information, identify endpoints, select paths, share capacity, detect or recover from errors, and regulate traffic.

Important performance dimensions include:

• Bandwidth or link rate: the nominal number of bits transmitted per second.
• Goodput: useful application data delivered per second, excluding protocol overhead and retransmitted data.
• Latency: time between sending and receiving an item or completing an operation.
• Jitter: variation in packet delay, important for voice, video, and industrial control.
• Loss rate: fraction of packets or information units not delivered successfully.
• Availability: probability that the service is operational when required.
• Scalability: ability to grow in users, traffic, or geographic scope without unacceptable degradation.

These measures can conflict. Increasing redundancy may improve availability but consume capacity. Deep buffering can reduce loss during short bursts but increase delay. Strong reliability mechanisms improve delivery probability but may reduce goodput under severe congestion. Network architecture is therefore a problem of constrained trade-offs rather than maximizing one metric.

2. Switching Techniques

2.1 Circuit Switching

Circuit switching reserves a communication path and resources for a session. A conventional circuit-switched connection has three phases: circuit establishment, data transfer, and teardown. Resources may be reserved in time slots, frequency bands, wavelengths, or physical channels.

If a connection requires setup time \(t_s\), sends \(L\) bits at rate \(R\), and experiences propagation delay \(t_p\), a simplified one-way completion time is

\[

T_{circuit}=t_s+\frac{L}{R}+t_p,

\]

assuming setup has already been represented as a one-time cost and ignoring processing and teardown. The model is appropriate for a dedicated connection, not for a general packet network.

Circuit switching offers predictable capacity, bounded contention after setup, and stable delay. Its weakness is inefficient utilization for bursty sources: reserved capacity remains idle during silence or inactivity. It is therefore well suited to constant-bit-rate streams and historically important in telephony.

2.2 Message Switching

In message switching, an entire message is stored at each intermediate node before being forwarded. No dedicated path is required. This store-and-forward method can use links efficiently but may require substantial memory and causes delay proportional to message size. If a message of length \(L\) traverses \(H\) equal-rate links, the transmission component is approximately \(HL/R\), in addition to propagation, processing, and queueing delays.

Message switching is unsuitable for interactive traffic when messages are large, because one damaged or delayed message may block substantial storage and forwarding capacity. Modern packet networks retain the store-and-forward principle but divide data into smaller packets.

2.3 Packet Switching

Packet switching divides a message into packets. In datagram networks, each packet is independently forwarded and may follow a different route. IP is the principal example. In virtual-circuit networks, a logical path is established, and packets generally follow that path, although resources may or may not be reserved.

Packet switching statistically multiplexes bursty sources. It improves aggregate utilization but introduces contention, variable queueing delay, packet loss, reordering, and sometimes retransmission. For a packet of \(L\) bits on a link of rate \(R\), transmission delay is

\[

t_{tx}=\frac{L}{R}.

\]

This is the time to place all bits on the link; it is not the time for the first bit to propagate to the receiver.

2.4 Comparative Evaluation

Circuit switching provides reservation and predictable service but may waste capacity. Message switching avoids reservation but has high storage and delay requirements. Packet switching is efficient for bursty traffic and scalable, but performance depends on traffic intensity, buffering, routing, and congestion control. Contemporary networks often combine approaches: packet transport may carry flows with priority, traffic engineering, admission control, or resource reservations.

3. Layered Architectures and Protocol Functions

3.1 Rationale for Layering

A protocol is a set of rules governing message formats, sequencing, interpretation, and actions. Layering decomposes a complex system into modules. Each layer offers services to the layer above and uses services of the layer below. A layer communicates logically with its peer at another node, but actual data moves through the local stack and physical medium.

Layering provides:

• modular implementation and testing;
• abstraction of lower-level details;
• interoperability through standardized interfaces;
• replacement of one technology without redesigning every other function;
• clearer fault isolation.

Its disadvantages include duplicated functions, processing overhead, rigid interfaces, and difficulty optimizing across layer boundaries.

3.2 OSI Reference Model

The seven-layer OSI model is a conceptual reference framework:

1. Physical: signals, connectors, modulation, bit timing, and transmission media.

2. Data link: framing, medium access, link-level error detection, and local delivery.

3. Network: logical addressing and forwarding across interconnected networks.

4. Transport: process-to-process delivery, reliability, flow control, and congestion control.

5. Session: dialog management and synchronization.

6. Presentation: syntax translation, representation, compression, and encryption concepts.

7. Application: network services visible to applications.

The OSI model is valuable pedagogically and analytically; it is not a complete mapping of every modern implementation.

3.3 TCP/IP Model

The practical Internet architecture is commonly described using four or five layers. A five-layer teaching model contains physical, link, network, transport, and application layers. The Internet Protocol provides best-effort network-layer datagram delivery. TCP provides reliable, ordered byte-stream transport with congestion control; UDP provides a connectionless datagram service with minimal transport functions. Application protocols include HTTP, DNS, SMTP, and many others.

OSI session and presentation functions are usually implemented within applications or libraries rather than as universal independent layers. The TCP/IP model is therefore an implementation-oriented architecture, whereas OSI is a more explicit reference decomposition.

3.4 Encapsulation and Demultiplexing

When an application sends data, each lower layer adds a control header and sometimes a trailer. Application data becomes a transport segment or datagram, which becomes a network packet, then a link-layer frame, and finally a physical bit stream. At the receiver, headers are removed in reverse order: decapsulation.

If application payload is \(P\) bytes, and each of four protocol layers adds header sizes \(h_1,h_2,h_3,h_4\), total transmitted size is

\[

P_{total}=P+\sum_{i=1}^{4}h_i.

\]

The overhead fraction relative to transmitted data is

\[

\eta_{overhead}=\frac{\sum h_i}{P+\sum h_i}.

\]

This assumes headers are the only overhead and excludes padding, retransmissions, and lower-layer coding. Small payloads suffer a greater relative overhead.

Demultiplexing uses identifiers to deliver received data to the correct protocol or application. Examples include EtherType at a link interface, an IP protocol field, and transport-layer port numbers. A port identifies an endpoint process or service within a host; it is not, by itself, a globally unique host address.

4. Addressing and Forwarding

Addressing occurs at several scopes:

• Names identify resources for human or application use, such as domain names.
• Network-layer addresses identify interfaces or logical locations for routing.
• Link-layer addresses support delivery over one local link.
• Transport ports identify application endpoints within a host.

A router forwards according to a destination network prefix and a forwarding table. It need not know the complete end-to-end path. Hierarchical addressing enables route aggregation, reducing table size and limiting update traffic. However, aggregation can produce suboptimal paths and may complicate traffic engineering.

Address resolution maps a network-layer destination to a local link-layer identifier. This mapping is local to a link or broadcast domain and should not be confused with end-to-end routing. Network address translation can conserve address space and provide a form of isolation, but it weakens transparent end-to-end reachability and can complicate peer-to-peer applications, diagnostics, and security protocols.

5. Multiplexing and Resource Sharing

Multiplexing combines multiple signals or flows over a shared resource; demultiplexing separates them at the receiver.

• Frequency-division multiplexing (FDM): users occupy different frequency bands, with guard bands to limit interference.
• Time-division multiplexing (TDM): users receive distinct time slots. Synchronous TDM may waste slots when a source is idle; statistical TDM assigns slots dynamically.
• Wavelength-division multiplexing (WDM): optical channels use different wavelengths on one fiber.
• Code-division techniques: users share time and frequency while using distinguishable spreading codes, subject to interference and power constraints.
• Orthogonal frequency-division multiplexing (OFDM): data uses many closely spaced orthogonal subcarriers, improving robustness against frequency-selective channels.

Packet statistical multiplexing is particularly effective when independent sources are bursty. Yet aggregation increases the probability that instantaneous demand exceeds link capacity. Buffers absorb short-term variation but cannot solve persistent overload.

6. Network Topologies and Architectures

A bus connects nodes to a shared medium; it is simple but vulnerable to contention and a single-medium failure. A star connects endpoints to a central switch; it is manageable and common in local networks, but the central device is critical. A ring connects each node to two neighbors and can offer deterministic access, although a break requires protection or reconfiguration. A tree provides hierarchical aggregation and scalability but can concentrate failures near the root. A mesh supplies path diversity and resilience at higher cost. Wireless networks frequently use infrastructure, mesh, ad hoc, or cellular topologies.

Topology is physical and logical. Two networks may have the same cable arrangement but different forwarding or overlay structures. Data-center fabrics often use highly connected leaf-spine designs to provide predictable multi-path connectivity, while wide-area networks use routers, optical systems, and engineered redundancy.

7. Service Models and QoS

A service model specifies what a network promises to an application. A best-effort model generally offers no strict guarantee of delivery, delay, bandwidth, or ordering. It is scalable and robust but requires applications to tolerate variation.

A reliable byte-stream service, such as TCP, hides packet loss and reordering through sequence numbers, acknowledgements, retransmission, and flow control. Reliability does not guarantee bounded delay: retransmission may improve eventual delivery while increasing latency. A datagram service, such as UDP, exposes loss and ordering behavior to the application and can support latency-sensitive or application-controlled recovery.

QoS is multidimensional. A useful traffic contract may specify peak rate, average rate, burst tolerance, maximum delay, jitter, loss probability, and availability. A token-bucket model characterizes an admissible arrival process by rate \(r\) and bucket size \(B\): over an interval of length \(t\), admitted traffic is bounded approximately by \(B+rt\), subject to the precise definition of the policing algorithm. This is a traffic constraint, not a guarantee that the network will provide the same rate.

Priority scheduling, weighted fair queuing, admission control, traffic shaping, policing, reservation, and differentiated services are tools for QoS. QoS mechanisms cannot create capacity; they allocate scarce capacity and determine which traffic experiences degradation during overload.

8. Delay, Throughput, and Queueing Analysis

8.1 Components of Delay

For a packet traversing a path, total delay can be modeled as

\[

T_{end-to-end}=T_{proc}+T_{queue}+T_{tx}+T_{prop},

\]

summed across links and nodes. Processing delay includes header examination and error checks. Queueing delay depends on current traffic. Transmission delay is \(L/R\). Propagation delay is

\[

t_{prop}=\frac{d}{v},

\]

where \(d\) is physical distance and \(v\) is signal propagation speed in the medium. Transmission and propagation are distinct: increasing link rate reduces transmission delay but does not materially reduce propagation delay.

For a path of \(H\) identical store-and-forward links, with no queueing and negligible processing, a packet of length \(L\) has approximately

\[

T=H\left(\frac{L}{R}+\frac{d}{v}\right).

\]

This expression assumes the packet must be fully received before forwarding and that each link has the same rate and distance. Cut-through forwarding or unequal links require a different model.

8.2 Little’s Law

For a stable system in long-run equilibrium,

\[

N=\lambda T,

\]

where \(N\) is the average number of items in the system, \(\lambda\) is the average arrival rate, and \(T\) is average time in the system. Stability requires that the long-run admitted arrival rate not exceed sustainable service capacity. Little’s Law is broadly applicable and does not require Poisson arrivals, but averages must refer to the same system boundary and time interval.

For example, if a router queue contains an average of 18 packets and admits 600 packets/s, average queueing-plus-service time is

\[

T=\frac{18}{600}=0.03\text{ s}=30\text{ ms}.

\]

This does not imply that every packet experiences 30 ms.

8.3 An M/M/1 Illustration

For an M/M/1 queue, arrivals are Poisson with rate \(\lambda\), service times are exponentially distributed with rate \(\mu\), and one server operates under a standard work-conserving discipline. The utilization is

\[

\rho=\frac{\lambda}{\mu}, \qquad 0\leq\rho<1.

\]

Under these assumptions,

\[

L=\frac{\rho}{1-\rho},\qquad W=\frac{1}{\mu-\lambda},

\]

where \(L\) is average number in the system and \(W\) is average time in the system. The average queue length and waiting time excluding service are

\[

L_q=\frac{\rho^2}{1-\rho},\qquad W_q=\frac{\rho}{\mu-\lambda}.

\]

These formulas must not be applied indiscriminately to real packet queues: packet arrivals may be correlated, packet sizes may vary, service may be scheduled among classes, and buffers may be finite. They are useful as a baseline and for demonstrating the nonlinear effect of utilization.

If \(\lambda=800\) packets/s and \(\mu=1000\) packets/s, then \(\rho=0.8\),

\[

W=\frac{1}{1000-800}=0.005\text{ s}=5\text{ ms},

\]

and

\[

L_q=\frac{0.8^2}{1-0.8}=3.2\text{ packets}.

\]

Increasing utilization to 0.95 gives \(W=1/(1000-950)=20\) ms, illustrating why delay rises sharply near capacity.

8.4 Bottlenecks and Goodput

For a flow crossing links with capacities \(R_1,\ldots,R_H\), the ideal steady-state throughput cannot exceed the bottleneck capacity

\[

R_{bottleneck}=\min_i R_i.

\]

Actual goodput may be lower because of headers, acknowledgements, contention, retransmissions, receiver limitations, or congestion control. A high-capacity access link cannot eliminate a lower-capacity core or server bottleneck.

9. End-to-End Design Principles

The end-to-end argument states that some functions are best implemented at communicating endpoints rather than only inside the network. Examples include application-level integrity checks, semantic correctness, encryption, and reliable transfer. Lower layers may still provide optimization and local recovery, but endpoint implementation is necessary when only endpoints possess the relevant application context.

Related principles include:

• Fate sharing: state whose loss should accompany the failure of the associated endpoint is often placed near that endpoint.
• Layered modularity: expose stable interfaces while allowing internal implementation change.
• Separation of control and data: routing and policy decisions can be distinct from high-rate forwarding.
• Robustness: tolerate malformed input, partial failure, delay, and reordering without unsafe behavior.
• Resource fairness: define whether fairness is per flow, user, service class, or economic stakeholder.
• Security by design: authenticate identities, protect confidentiality and integrity, minimize privilege, and plan key management.

The end-to-end principle is not an absolute ban on network functionality. Firewalls, congestion control, multicast support, in-network caching, and local error recovery may be justified by performance, security, or operational requirements. The design question is where a function can be implemented correctly, efficiently, and with acceptable trust assumptions.

10. Standards, Governance, and Professional Practice

The Internet Engineering Task Force develops and documents Internet protocols through Requests for Comments and an open technical process. The IEEE standardizes many local and metropolitan networking technologies, including Ethernet and wireless LAN families. The International Telecommunication Union develops international telecommunications recommendations and spectrum-related frameworks. ISO and IEC publish broad information-technology and systems standards. The World Wide Web Consortium develops Web standards.

Standards improve interoperability but do not guarantee identical implementations. Engineers must distinguish mandatory requirements from recommendations, version dependencies, profiles, and vendor extensions. Regulatory compliance may involve spectrum licensing, lawful-intercept rules, privacy and data-protection requirements, accessibility, export controls, and critical-infrastructure obligations. Network telemetry and packet inspection should observe data minimization, authorization, retention, and transparency principles.

11. Contemporary Developments and Research Directions

Software-defined networking separates or logically centralizes control functions from programmable forwarding, enabling policy automation and traffic engineering. Network-function virtualization moves functions such as firewalls and load balancers into software, improving flexibility but introducing orchestration and performance challenges. Edge computing reduces application latency by placing computation near users or sensors. Content delivery networks cache content near demand points, reducing backbone traffic.

Artificial intelligence can support anomaly detection, traffic prediction, routing optimization, and capacity planning. However, training-data bias, adversarial manipulation, opaque decisions, privacy exposure, and unstable behavior under distribution shift require human oversight and measurable safeguards. AI should augment validated control processes rather than replace safety-critical engineering without evidence.

Emerging networks combine deterministic networking, time-sensitive industrial communication, programmable data planes, satellite and non-terrestrial access, and advanced optical systems. Research questions include dependable low-latency service, energy-aware networking, secure multi-domain orchestration, quantum-resistant security, and jointly optimized communication and computation.

12. Illustrative Case: QoS Design for a Remote Industrial Monitoring Network

Background

A manufacturing plant connects vibration sensors and video cameras to an edge controller. Sensors generate periodic small reports, while cameras produce bursty high-rate traffic. The control application requires low delay and high delivery probability; video can tolerate moderate delay and occasional loss.

Problem

A single 100 Mb/s uplink carries an average sensor load of 8 Mb/s and an average video load of 70 Mb/s. During production events, video rises to 95 Mb/s. Protocol overhead, acknowledgements, and bursts cause queue growth, even though the long-run average is below 100 Mb/s.

Analysis

The peak offered load during an event is at least 103 Mb/s before overhead, exceeding the link rate. No scheduling algorithm can transmit all traffic immediately. Best-effort FIFO service allows video bursts to delay control packets. Large buffers prevent immediate loss but can create unacceptable control latency. The correct conclusion is not simply to assign higher priority; the bottleneck must also be addressed.

Approach

The plant implements: (1) a separate high-priority class for control packets; (2) traffic policing and shaping for video; (3) admission control for new cameras; (4) a larger or additional uplink; (5) redundant paths for availability; and (6) measurements of one-way delay, jitter, loss, queue occupancy, and achieved goodput. Control packets receive a bounded service opportunity, while video is shaped to its contracted rate. Encryption and authenticated device identities protect telemetry.

Outcome

Priority scheduling reduces control delay during video bursts, but capacity expansion is required to prevent persistent overload. Monitoring confirms whether the design meets the control application's measured delay and loss objectives rather than relying on nominal link rate.

Learning Points

QoS is a system property, not merely a packet-marking operation. Traffic contracts must reflect peaks and bursts, queues trade loss for delay, and end-to-end measurements are necessary because local link statistics cannot establish application performance.

Challenges and Limitations

Network models simplify reality. Deterministic delay bounds require assumptions about traffic, scheduling, and failure behavior that may not hold in shared public networks. Queueing models may misrepresent self-similar or correlated traffic. Layer boundaries can obscure cross-layer interactions, while centralized controllers introduce control-plane scalability and failure concerns. Security mechanisms consume computational and bandwidth resources. Finally, interoperability may be limited by proprietary extensions even when nominal standards are shared.

Best-Practice Design Checklist

1. State application requirements in measurable terms: rate, burst size, delay, jitter, loss, availability, and security.

2. Identify bottlenecks and failure domains before selecting protocols.

3. Distinguish nominal capacity from goodput.

4. Model both average and peak traffic.

5. Validate queueing assumptions and use measurements to calibrate models.

6. Use hierarchical addressing and route aggregation where appropriate.

7. Select reliability and QoS mechanisms according to application semantics.

8. Apply defense in depth, least privilege, secure management, and data minimization.

9. Test interoperability, overload, failure recovery, and reconfiguration.

10. Document standards versions, assumptions, and service-level objectives.

Chapter Summary

Communication networks are layered systems that transport information through shared resources. Circuit switching reserves resources and offers predictable service; message switching stores complete messages; packet switching statistically multiplexes smaller units and scales efficiently for bursty traffic. OSI provides a seven-layer reference model, while TCP/IP reflects practical Internet organization. Encapsulation adds headers as data descends the stack, and demultiplexing uses addresses, protocol identifiers, and ports to deliver data correctly.

Multiplexing and topology determine how resources and paths are shared. Service models range from best effort to reliable or QoS-controlled delivery, but reliability does not itself guarantee bounded delay. End-to-end performance depends on processing, queueing, transmission, and propagation components. Little’s Law provides a general relationship among average population, rate, and time; M/M/1 results provide a useful but assumption-dependent queueing baseline. Utilization near unity produces sharply increasing delay.

Sound architecture combines modularity with end-to-end correctness, security, resilience, measurable QoS, and awareness of bottlenecks. Standards bodies enable interoperability, while contemporary systems increasingly use programmable control, virtualization, edge computing, and AI-assisted operations. Effective network design therefore integrates protocol theory, quantitative analysis, governance, and empirical validation.

Key Terms

Protocol: Rules governing communication syntax, semantics, and behavior.

Layering: Decomposition of network functions into cooperating modules.

Encapsulation: Addition of control information by lower protocol layers.

Decapsulation: Removal and interpretation of protocol headers at reception.

Circuit switching: Communication using a reserved logical or physical circuit.

Message switching: Store-and-forward transfer of complete messages.

Packet switching: Transfer of messages divided into independently handled packets.

Multiplexing: Combining multiple signals or flows over a shared resource.

Demultiplexing: Separating a combined stream into its constituent flows.

Goodput: Useful application data delivered per unit time.

Propagation delay: Time for a signal to travel through a medium.

Transmission delay: Time required to place a packet’s bits on a link.

Queueing delay: Waiting time caused by contention for a resource.

Utilization: Fraction of service capacity being used.

QoS: Measurable service characteristics such as delay, loss, and rate.

Little’s Law: Stable-system relationship \(N=\lambda T\).

Bottleneck: Resource limiting achievable end-to-end throughput.

End-to-end principle: Placement of correctness functions at communicating endpoints when endpoint context is necessary.

Traffic shaping: Controlling the timing or rate of transmitted traffic.

Route aggregation: Combining address prefixes to reduce routing information.

Review Questions

Short Answer Questions

1. What are the principal differences between circuit and packet switching?

2. Why are transmission delay and propagation delay distinct?

3. Explain encapsulation and decapsulation.

4. What is the role of a transport-layer port?

5. State Little’s Law and define its variables.

Long Answer Questions

1. Compare OSI and TCP/IP architectures, including their purposes and limitations.

2. Explain FDM, TDM, WDM, and statistical multiplexing with suitable applications.

3. Discuss best-effort, reliable-stream, and datagram service models.

4. Derive and interpret the M/M/1 expressions for utilization, average system time, and queue length under their assumptions.

5. Explain how the end-to-end principle influences reliability, security, and network-layer design.

Analytical / Application Questions

1. A 1200-byte packet crosses four 10 Mb/s links, each 500 km long. Assuming store-and-forward, negligible processing and queueing, and propagation speed \(2\times10^8\) m/s, calculate the total transmission-plus-propagation delay.

2. A router queue has an average of 24 packets and an admitted rate of 800 packets/s. Use Little’s Law to calculate average time in the system, then explain why this average is not a bound for every packet.

3. A 50 Mb/s link carries a 15 Mb/s sensor stream and bursty video whose instantaneous demand reaches 45 Mb/s. Evaluate the overload condition and propose two technically distinct remedies, discussing their trade-offs.

Multiple Choice Questions with Answers

Question 1

Which switching method reserves resources for a connection before data transfer?

A. Circuit switching

B. Datagram packet switching

C. Message switching

D. Best-effort forwarding

Correct Answer: A. Circuit switching

Question 2

What does transmission delay L/R represent?

A. Time for a signal to propagate through a medium

B. Time to place L bits on a link of rate R

C. Time spent waiting in a router queue

D. Time required to establish a route

Correct Answer: B. Time to place L bits on a link of rate R

Question 3

Which OSI layer is primarily responsible for logical addressing and forwarding across interconnected networks?

A. Physical

B. Data-link

C. Network

D. Presentation

Correct Answer: C. Network

Question 4

A payload is 1000 bytes and four headers add 20, 8, 20, and 14 bytes. What is the transmitted size?

A. 1000 bytes

B. 1048 bytes

C. 1054 bytes

D. 1062 bytes

Correct Answer: D. 1062 bytes

Question 5

Which statement about Little’s Law is correct for a stable system?

A. N = λT

B. N = λ/T

C. T = Nλ

D. λ = NT

Correct Answer: A. N = λT

Question 6

For an M/M/1 queue, which condition is required for a finite steady-state average delay?

A. λ = 0

B. λ > μ

C. λ < μ

D. λ = μ

Correct Answer: C. λ < μ

Question 7

Which multiplexing method assigns distinct optical wavelengths to channels on one fiber?

A. TDM

B. WDM

C. FDM in copper only

D. Code division only

Correct Answer: B. WDM

Question 8

A link has capacity 100 Mb/s, while a flow crosses another link of capacity 40 Mb/s. Ignoring other limitations, what is the maximum bottleneck-limited throughput?

A. 40 Mb/s

B. 60 Mb/s

C. 100 Mb/s

D. 140 Mb/s

Correct Answer: A. 40 Mb/s

Question 9

Which statement best reflects the end-to-end principle?

A. All reliability must be implemented only in routers

B. Functions requiring application context are often implemented at endpoints

C. Physical-layer coding is unnecessary

D. Network-layer forwarding is performed only by applications

Correct Answer: B. Functions requiring application context are often implemented at endpoints

Question 10

A 100 Mb/s link has 5% overhead from each of five independent layers, modeled as a total subtraction of 5 × 12 Mb/s. What is the resulting rate under this stated model?

A. 25 Mb/s

B. 40 Mb/s

C. 60 Mb/s

D. 95 Mb/s

Correct Answer: B. 40 Mb/s