Mobility, Routing and End-to-End Performance

Introduction

Wireless networks differ from fixed networks because users move, links vary, radio resources are shared, and packet loss may result from fading rather than congestion. These conditions challenge the Internet's traditional assumptions: an IP address identifies a stable attachment point, paths change slowly, and loss usually indicates overload. This chapter develops the mechanisms used to maintain reachability and performance as devices move. It covers mobility management, location tracking, Mobile IPv4, wireless routing, transport-layer behavior, congestion control, quality of service (QoS), and cross-layer design.

The central design problem is to preserve communication continuity while simultaneously managing changing location, topology, capacity, delay, interference, and energy constraints.

Learning Objectives

After studying this chapter, a student should be able to:

1. Explain location management, handover, and mobility-management architectures.
2. Describe Mobile IPv4 packet delivery, care-of addresses, tunneling, and triangular routing.
3. Compare proactive, reactive, and hybrid wireless-routing protocols.
4. Analyze how wireless loss, delay, and variability affect TCP and other transport protocols.
5. Apply basic throughput, delay, and QoS calculations.
6. Explain congestion-control mechanisms and distinguish congestion loss from wireless loss.
7. Evaluate cross-layer strategies and their limitations.

1. Mobility Management and Location Tracking

1.1 Types of mobility

A mobile node changes its point of network attachment while maintaining communication. Mobility can be classified as:

• Nomadic mobility: the device changes location between sessions but does not require session continuity.
• Continuity mobility: active sessions survive movement between access points or subnets.
• Micromobility: movement within one administrative domain, often handled locally.
• Macromobility: movement between domains, operators, or wide-area networks.

A handover (or handoff) consists of discovering a target access point, authenticating or authorizing the device, allocating resources, updating forwarding state, and transferring data. In a hard handover, the old link is released before the new link is established. In a soft or make-before-break procedure, both links may temporarily exist, reducing interruption at the cost of extra resources.

1.2 Location-management cycle

Location management has two complementary operations:

1. Location update: the mobile device or network reports a new location.
2. Paging or lookup: the network searches the last known area when traffic arrives.

If a device updates after every movement, lookup is inexpensive but signaling is high. If it updates only rarely, signaling is reduced but paging becomes expensive. A simplified cost model is

where C_total is total signaling cost, C_u is cost per update, N_u is the number of updates, C_p is cost per paging operation, and N_p is the number of paging operations. This is a planning model, not a universal mobility law; actual costs depend on topology, timers, radio conditions, and user behavior.

Location areas, tracking areas, registration zones, and visitor databases are examples of mechanisms that aggregate several cells or access points. Aggregation reduces update frequency but increases the area that must be searched.

1.3 Handover decision factors

A handover should not be triggered solely by instantaneous received signal strength. Practical algorithms consider signal quality, interference, velocity, load, predicted residence time, authentication delay, and available capacity. A hysteresis margin prevents repeated switching between two access points. A time-to-trigger condition requires a measurement to remain below or above a threshold for a specified interval.

The handover interruption time may be approximated as

where the terms represent discovery, authentication, configuration, signaling-update, and buffering contributions. This additive model is useful for budgeting, although some operations may overlap in an implementation.

2. Mobile IP

2.1 The addressing problem

Conventional IP routing assumes that an address identifies a topological location. If a node retains its home address after moving to another network, ordinary routers still send packets toward the home network. If the node changes its address, ongoing transport connections may fail because the endpoint address changes.

Mobile IP separates a stable identity from a temporary point of attachment. The principal entities in Mobile IPv4 are:

• Mobile node (MN): the moving device.
• Home agent (HA): a router on the home network that intercepts packets for the MN and tunnels them toward its current care-of address.
• Foreign agent (FA): a router on the visited network that may provide a care-of address and decapsulate tunneled packets.
• Correspondent node (CN): the communication peer.

2.2 Care-of addresses and registration

A care-of address (CoA) identifies the MN's current attachment point. Two forms are important:

1. Foreign-agent care-of address: the FA owns or advertises the address. The HA tunnels packets to the FA. The FA normally decapsulates the packets and delivers the original datagrams to the MN on the visited link.
2. Co-located care-of address: the MN acquires an address, for example through DHCP. The HA tunnels packets directly to that address, and the MN itself decapsulates the packets.

This distinction is essential: the mobile node does not always perform decapsulation.

The MN discovers an available agent or obtains a co-located address, sends a registration request, and receives a registration reply. Registration establishes a binding between the home address and CoA. Authentication prevents unauthorized redirection of traffic.

2.3 Packet processing

For a packet sent from CN to MN, the basic path is:

1. CN sends an IP datagram to the MN's home address.
2. The HA intercepts the datagram on the home network.
3. The HA encapsulates the original datagram inside a new IP packet addressed to the CoA.
4. With an FA CoA, the FA decapsulates the outer packet and forwards the original datagram to the MN. With a co-located CoA, the MN decapsulates it.
5. The MN processes the original packet using its home address.

This process is called IP-in-IP tunneling. The outer header provides routing across the visited network; the inner header preserves the original endpoints.

A common reverse path is triangular routing: CN → HA → FA/MN in the forward direction, but MN → CN directly in the reverse direction. Reverse tunneling can send packets from MN through the FA and HA when ingress filtering or policy requires a source address associated with the home network. Route optimization can reduce path stretch, but it requires secure distribution and maintenance of mobility bindings.

2.4 Encapsulation overhead and fragmentation

If an outer IP header of 20 bytes is added to an original packet, the tunneled packet grows by 20 bytes, excluding optional headers. If the original payload and headers total 1,480 bytes, the encapsulated size is 1,500 bytes. A link MTU of 1,500 bytes can therefore carry the packet only if no additional tunnel header is required; otherwise fragmentation, path-MTU discovery, or a smaller packet is needed.

Tunneling increases overhead and may increase delay. It can also create a nonoptimal route and a single point of failure at the home agent. Modern cellular systems generally use different mobility architectures, including anchor-based and distributed mobility, but the Mobile IP model remains foundational.

3. Wireless Routing

Wireless routing protocols must handle changing links, limited bandwidth, broadcast transmission, hidden terminals, asymmetric links, energy constraints, and unreliable measurements. A route that is shortest in hop count may be poor if it uses congested or unstable links.

3.1 Routing metrics

Possible metrics include hop count, expected transmission count (ETX), delay, residual energy, interference, link capacity, and reliability. A path metric may be represented generally as

where m_i is a normalized metric for hop i, w_i is its weight, and k is the number of hops. Weighted sums are design choices; they do not automatically produce an optimal route unless the weights and measurements reflect the application's objective.

For a link with forward delivery probability d_f and reverse delivery probability d_r, a commonly used ETX estimate is

assuming independent attempts and stable probability estimates. It estimates expected transmissions, including retransmissions, for successful data and acknowledgment exchange. It is not a direct measure of delay when contention and queuing dominate.

3.2 Protocol families

Proactive protocols maintain routes continuously. Destination-Sequenced Distance Vector (DSDV) uses sequence numbers to reduce loops. Proactive operation offers low data-start delay but generates periodic control traffic.

Reactive protocols discover routes only when needed. Ad hoc On-Demand Distance Vector (AODV) broadcasts a route request, receives a route reply, and uses sequence numbers and route errors. Dynamic Source Routing (DSR) carries the complete route in packet headers, which can become expensive as paths grow.

Hybrid protocols combine local proactive information with on-demand discovery beyond a zone. They attempt to balance control overhead and route-discovery delay.

Geographic routing forwards using node positions, often selecting a neighbor that makes progress toward the destination. It reduces route-state maintenance but requires location information and can fail at geographic voids.

3.3 Routing design trade-offs

A route-discovery broadcast consumes shared wireless capacity and may collide with data traffic. Frequent route repair can cause packet reordering and bursts of retransmission. Multipath routing improves resilience but may create duplicate packets and out-of-order delivery. Energy-aware routing can prolong network life but may deliberately choose longer paths.

A suitable protocol depends on mobility, node density, traffic pattern, energy budget, security requirements, and administrative control. No routing family is universally best.

4. End-to-End Performance

4.1 Delay, loss, and throughput

End-to-end delay can be decomposed as

where propagation, transmission, processing, queuing, and handover terms are measured along the complete path. In a multihop network, each hop contributes its own propagation, transmission, processing, and queuing delay.

If a packet of length L bits is transmitted over a link of rate R bits/s, the serialization time is

For a long-lived flow, average goodput is successfully delivered application data divided by elapsed time. It should be distinguished from link rate, which excludes many protocol and retransmission effects.

4.2 TCP over wireless paths

TCP interprets duplicate acknowledgments, timeouts, and missing acknowledgments primarily as signs of congestion. Wireless losses caused by fading, interference, route breaks, or handovers can therefore trigger unnecessary congestion-window reduction. Variable round-trip time can also cause premature retransmission timeouts.

TCP congestion control commonly includes:

• Slow start: congestion window grows approximately exponentially per round-trip time until a threshold or loss event.
• Congestion avoidance: window grows approximately linearly.
• Fast retransmit: duplicate acknowledgments prompt retransmission before a timeout.
• Fast recovery: the sender reduces its sending rate but may avoid returning to the smallest initial window.

The standard retransmission timeout is based on a smoothed RTT and its variation. One representative estimator is

where 0 < α < 1, RTTVAR is a smoothed deviation measure, and implementations also apply lower and upper bounds and exponential backoff. Exact update order and constants are protocol-specific.

4.3 Mathis throughput approximation

For a long-lived TCP flow with congestion avoidance, a relatively stable RTT, independent random loss probability p, and sufficiently large buffers, the Mathis approximation is often written as

Here T_bytes is approximate throughput in bytes per second, MSS is maximum segment size in bytes, RTT is seconds, and p is a dimensionless loss probability. The approximation hides constants associated with the congestion-control algorithm; it is not a universal exact law.

If throughput is required in bits per second, use

For MSS = 1,000 bytes, RTT = 0.1 s, and p = 0.01,

√p = 0.1 and therefore T_bytes ≈ 1000/(0.1 × 0.1) = 100,000 bytes/s, or T_bits ≈ 800,000 bits/s. This example is valid only as an approximation under the stated assumptions; handovers, burst losses, delayed acknowledgments, receiver windows, and radio scheduling can produce very different results.

4.4 Improving transport performance

Solutions include link-layer local retransmission, explicit loss notification, split connections, proxy-based transport, multipath transport, forward-error correction, and improved handover signaling. Each has costs. Local recovery can hide losses but may increase delay; split connections break end-to-end security and semantics; aggressive redundancy consumes capacity; multipath scheduling must manage reordering and path heterogeneity.

5. Congestion Control and QoS

5.1 Congestion versus wireless loss

Congestion occurs when offered traffic exceeds available service capacity, causing queues to grow and packets to be delayed or discarded. Wireless loss can occur even when queues are empty. Measurements should therefore consider queue occupancy, channel utilization, retransmission counts, signal quality, route changes, and delay variation rather than treating every loss as congestion.

A useful stability condition for a queue is

where λ is average arrival rate and μ is average service rate, both in packets/s. If ρ ≥ 1 for a sustained period in a simple single-server model, the queue cannot remain stable. Real wireless systems have variable service rates, so this is a conceptual condition rather than a complete performance model.

5.2 QoS dimensions

QoS includes throughput, delay, jitter, loss probability, reliability, availability, and fairness. Applications have different requirements: voice is sensitive to delay and jitter, industrial control may require bounded latency and high reliability, and file transfer generally tolerates delay but seeks high goodput.

A QoS service specification should define measurable bounds, traffic assumptions, admission rules, and treatment during overload. Priority alone is not QoS: without admission control, high-priority flows can still overload the system.

Useful mechanisms include traffic shaping, policing, scheduling, admission control, reservation, differentiated queues, airtime fairness, and application adaptation. Wireless QoS must account for contention and channel variation because a reserved queue does not guarantee a reserved radio channel.

6. Cross-Layer Design

Traditional layering provides modularity and allows independent evolution. However, wireless performance sometimes improves when information crosses layers. Examples include:

• The routing layer selects links using physical-layer quality estimates.
• The transport layer receives a handover or route-break indication.
• The MAC layer schedules packets according to application deadlines.
• The application adapts video rate to measured capacity and delay.
• The receiver provides channel-state information to guide coding or retransmission.

Cross-layer designs should use explicit interfaces, stable abstractions, and controlled feedback. Unrestricted coupling can create oscillations, duplicate functions, security vulnerabilities, and difficult testing. For example, a routing metric based on rapidly fluctuating signal strength may cause route instability; smoothing improves stability but makes the metric less responsive.

A practical design process is: define the application objective, identify measurable variables, specify the feedback timescale, bound the control action, test interactions under mobility and overload, and provide safe fallback behavior.

7. Applied Case: Mobility in a University Campus Network

Background

This illustrative case concerns a campus network supporting voice calls, learning-management traffic, and live video for students moving among buildings. Access points are connected to an IP core, while some outdoor areas use a wireless mesh.

Problem

Students report video freezes near building entrances. Monitoring shows that signal strength remains acceptable, but handovers take 500 ms, route discovery occasionally adds delay, and TCP reduces its congestion window after bursts of loss. Voice calls experience jitter during the same events.

Analysis

The problem is not simply weak signal. The handover procedure includes authentication and address configuration, while mesh route repair competes with application traffic. TCP interprets the resulting missing acknowledgments as congestion. The campus also uses a single best-effort queue, so video bursts affect voice packets.

Approach

The engineering team implements pre-authentication for likely target access points, buffers packets briefly during handover, and maintains local routes within each building. It applies a stable routing metric that combines ETX and measured delay rather than instantaneous signal strength. Voice receives a bounded-rate priority queue with admission control, while video uses an adaptive bitrate. Transport monitoring distinguishes route-change events from persistent queue growth.

Outcome

Handover interruption decreases, route-repair traffic is localized, and voice jitter falls. Video quality improves because its bitrate follows available capacity instead of repeatedly filling queues. The solution does not eliminate wireless loss; it reduces avoidable control delay and prevents one application's bursts from dominating the shared medium.

Learning points

1. Mobility performance is an end-to-end property.
2. Signal strength alone is an inadequate handover metric.
3. Localizing mobility control can reduce signaling and route disruption.
4. QoS requires admission and scheduling, not merely priority labels.
5. Cross-layer information is useful when interfaces and feedback are carefully controlled.

8. Security, Governance, and Contemporary Developments

Mobility signaling must be authenticated because a false binding can redirect traffic or enable interception. Wireless routing is vulnerable to fabricated identities, selective forwarding, wormholes, replay, and route poisoning. Protection requires cryptographic authentication, secure key management, replay protection, monitoring, and policy enforcement.

Privacy is also important. Persistent location updates can reveal movement patterns. Systems should minimize retained location data, restrict access, publish retention policies, and obtain appropriate consent. Regulatory obligations may govern lawful interception, emergency services, spectrum use, and protection of personal data.

Contemporary networks increasingly use 5G network slicing, edge computing, multipath transport, software-defined control, machine-learning-assisted radio optimization, and integrated access and backhaul. AI can predict mobility or traffic demand, but training data may be biased, predictions can fail under unusual mobility, and opaque decisions complicate accountability. AI should therefore be evaluated against safety constraints, explainability needs, privacy requirements, and deterministic fallback procedures.

Best-Practice Checklist

• Separate identity, location, and forwarding state.
• Distinguish foreign-agent and co-located Mobile IPv4 care-of addresses.
• Measure delay, jitter, loss cause, and route stability—not signal strength alone.
• Select routing metrics appropriate to the application and timescale.
• Treat Mathis-style throughput equations as approximations with explicit units.
• Use local recovery carefully so it does not conceal persistent congestion.
• Apply QoS with admission control and measurable service objectives.
• Design cross-layer feedback with bounded, stable control loops.
• Authenticate mobility and routing messages.
• Minimize collection and retention of location information.

Chapter Summary

Mobility requires coordinated location tracking, handover control, address management, and routing adaptation. Mobile IPv4 preserves a mobile node's home address by binding it to a care-of address and tunneling packets through the home agent. With a foreign-agent care-of address, the foreign agent normally decapsulates packets; with a co-located care-of address, the mobile node decapsulates them. Wireless routing must address unstable links, contention, energy limits, and changing topology, so hop count alone is often inadequate. TCP can mistake wireless loss and handover disruption for congestion, while the Mathis approximation relates long-lived throughput to MSS, RTT, and loss probability with carefully defined units. QoS requires control of delay, jitter, loss, capacity, and fairness under variable radio conditions. Cross-layer design can improve responsiveness but must preserve modularity, security, stability, and testability. Effective engineering combines mobility-aware signaling, appropriate routing metrics, transport adaptation, QoS mechanisms, and privacy-conscious governance.

Key Terms

Mobile node: A device that changes its network attachment point.

Handover: Transfer of an active connection from one access point or network to another.

Location update: Signaling that records a mobile device's current area or attachment.

Paging: Searching one or more areas to locate a mobile device.

Home agent: Mobile IPv4 router that maintains a binding for a mobile node.

Foreign agent: Visited-network router that may provide a care-of address and decapsulate packets.

Care-of address: Temporary address representing a mobile node's current attachment.

Co-located care-of address: Care-of address obtained and used directly by the mobile node.

Tunneling: Encapsulating an original packet inside another packet for delivery.

Triangular routing: Forward path through the home agent and reverse path directly to the correspondent.

Proactive routing: Routing that maintains paths before they are requested.

Reactive routing: Routing that discovers paths when traffic requires them.

ETX: Expected number of transmissions needed for successful delivery.

RTT: Round-trip time between sender and receiver.

MSS: Maximum segment size, measured in bytes in the throughput approximation used here.

Congestion control: Regulation of sending rate to avoid overload.

QoS: Measurable service properties such as delay, jitter, loss, and throughput.

Cross-layer design: Deliberate exchange of information between protocol layers.

Admission control: Decision process determining whether a new flow may receive requested service.

Review Questions

Short Answer Questions

1. Distinguish nomadic mobility from continuity mobility.
2. What are the two basic operations of location management?
3. Explain the difference between a foreign-agent and a co-located care-of address.
4. Why can TCP reduce its sending rate unnecessarily on a wireless path?
5. Name four dimensions of QoS.

Long Answer Questions

1. Explain Mobile IPv4 registration and packet forwarding from a correspondent node to a mobile node.
2. Compare proactive, reactive, hybrid, and geographic wireless-routing approaches.
3. Discuss how handover delay, route repair, and retransmission interact to affect end-to-end performance.
4. Explain congestion control and distinguish congestion loss from fading- or handover-induced loss.
5. Evaluate the benefits and risks of cross-layer design in a mobile wireless network.

Analytical / Application Questions

1. A tunnel adds a 20-byte outer IP header to a 1,480-byte packet. Determine the encapsulated size and discuss whether it fits a 1,500-byte MTU when no other tunnel header is present.
2. Using MSS = 1,200 bytes, RTT = 0.2 s, and p = 0.0025, calculate the Mathis approximate throughput in bytes/s and bits/s. State why the result is only an approximation.
3. Design a mobility and QoS strategy for a campus voice-and-video network. Identify suitable handover triggers, routing metrics, transport responses, and security controls.

Multiple Choice Questions with Answers

Question 1

With a foreign-agent care-of address in Mobile IPv4, who normally decapsulates the tunneled packet?

A. The correspondent node

B. The home agent

C. The foreign agent

D. The mobile node in every case

Correct Answer: C. The foreign agent

Question 2

A co-located care-of address is best described as an address that is:

A. Owned only by the home agent

B. Acquired and used directly by the mobile node

C. Assigned permanently to the correspondent node

D. Used only for broadcast routing

Correct Answer: B. Acquired and used directly by the mobile node

Question 3

In the approximation T_bytes ≈ MSS/(RTT√p), MSS is measured in:

A. Bytes

B. Bits

C. Packets per second

D. Seconds

Correct Answer: A. Bytes

Question 4

Which routing family discovers a route only when traffic requires one?

A. Proactive

B. Reactive

C. Circuit-switched

D. Static-only

Correct Answer: B. Reactive

Question 5

If the forward and reverse delivery probabilities are both 0.8, the ETX estimate 1/(df dr) is:

A. 0.64

B. 0.80

C. 1.25

D. 1.5625

Correct Answer: D. 1.5625

Question 6

Which event most directly indicates a handover interruption component?

A. Propagation delay on a fixed link

B. Authentication and address reconfiguration

C. The MSS value

D. The home address length

Correct Answer: B. Authentication and address reconfiguration

Question 7

For a simple queue model, the condition ρ = λ/μ < 1 indicates:

A. A stable average arrival-service relationship

B. Zero packet loss under all conditions

C. Infinite service delay

D. That arrival rate is exactly equal to service rate

Correct Answer: A. A stable average arrival-service relationship

Question 8

Which mechanism limits the admission of new flows to protect existing QoS commitments?

A. Route poisoning

B. Admission control

C. Packet fragmentation

D. Address translation

Correct Answer: B. Admission control

Question 9

For MSS = 1,000 bytes, RTT = 0.1 s, and p = 0.01, the Mathis approximation in bytes/s is:

A. 1,000

B. 10,000

C. 100,000

D. 800,000

Correct Answer: C. 100,000

Question 10

A major risk of unrestricted cross-layer coupling is:

A. Guaranteed elimination of congestion

B. Removal of all security requirements

C. Control instability and difficult testing

D. Permanent increase in MTU

Correct Answer: C. Control instability and difficult testing