Routing, Switching and Network Optimization

Introduction

Advanced communication networks must deliver packets across large, heterogeneous and continually changing topologies. Routing determines the logical path through an internetwork; switching forwards packets or frames at each node; and optimization selects paths and resource allocations that satisfy performance, reliability, policy and scalability requirements. These functions are interdependent. A routing protocol may compute a shortest path, but forwarding hardware must install that path, traffic engineering may deliberately select a non-shortest route, and policy may prohibit the technically shortest route altogether.

This chapter examines distance-vector, link-state, path-vector, multicast and traffic-engineering methods, with particular attention to OSPF, IS-IS, BGP and segment routing. It also develops graph-theoretic and constrained optimization models. A central distinction is made between feasibility—whether a route satisfies all constraints—and optimality—whether it is the best feasible route under a selected objective.

Learning Objectives

After completing this chapter, a learner should be able to:

1. Compare distance-vector, link-state and path-vector routing.

2. Explain the operation and design roles of OSPF, IS-IS and BGP.

3. Distinguish unicast, multicast and traffic-engineered forwarding.

4. Formulate routing as a graph and constrained optimization problem.

5. Apply reliability transformations correctly in path selection.

6. Explain segment routing and its relationship to centralized optimization.

7. Evaluate scalability, convergence, policy, security and operational trade-offs.

1. Network Graphs and Routing Decisions

A communication network can be represented by a directed or undirected graph

\[

G=(V,E),

\]

where \(V\) is the set of routers or switching nodes and \(E\) is the set of links. Each edge \(e\) may have attributes such as administrative cost \(c_e\), propagation delay \(d_e\), capacity \(u_e\), loss probability \(p_e\), and reliability \(r_e=1-p_e\), where \(0<r_e\leq 1\).

A path \(P\) from source \(s\) to destination \(t\) is an ordered sequence of adjacent edges. For additive metrics, such as cost or delay under an appropriate approximation,

\[

C(P)=\sum_{e\in P}c_e,

\qquad D(P)=\sum_{e\in P}d_e.

\]

The shortest-path problem is to find a feasible path minimizing \(C(P)\). Dijkstra’s algorithm is applicable when all edge costs are non-negative. Bellman–Ford-type relaxation can accommodate negative edge weights, but negative routing metrics are generally inappropriate in operational IP networks because they can create unstable or undefined shortest paths.

Reliability is not additive. If link failures are statistically independent and every link on a path must operate, path reliability is

\[

R(P)=\prod_{e\in P}r_e.

\]

Taking negative logarithms converts the multiplicative expression into an additive one:

\[

-\log R(P)=\sum_{e\in P}[-\log(r_e)].

\]

Thus, minimizing \(\sum -\log(r_e)\) maximizes path reliability, subject to the independence model. This transformation is not valid without qualification when failures are correlated; shared ducts, power systems or physical-risk domains can invalidate the independent-link assumption.

Worked Example 1: Cost and Reliability

Suppose two candidate paths have the following independent link reliabilities:

\[

P_1: 0.99\times0.98=0.9702,

\]

\[

P_2: 0.995\times0.995\times0.995=0.985074875.

\]

Although \(P_2\) uses three links rather than two, it is more reliable. A decision based only on hop count would therefore be incorrect. If the objective is reliability, the corresponding transformed costs are \(-\log(0.99), -\log(0.98)\), and so forth.

2. Routing Paradigms

2.1 Distance-Vector Routing

In distance-vector routing, each router maintains an estimate of the distance to destinations and periodically or event-triggeredly shares that estimate with neighbors. The Bellman optimality relation is

\[

D_x(y)=\min_{v\in N(x)}\{c(x,v)+D_v(y)\},

\]

where \(D_x(y)\) is router \(x\)'s estimate of the distance to destination \(y\), \(N(x)\) is the neighbor set, and \(c(x,v)\) is the link cost.

Advantages include modest memory requirements and a relatively simple implementation. Limitations include slow convergence, routing loops and the count-to-infinity problem. Split horizon, poisoned reverse, triggered updates and bounded metrics reduce—but do not universally eliminate—these problems. RIP is a classical distance-vector protocol and uses hop count as its principal metric, with a maximum finite hop-count limit that restricts network diameter.

2.2 Link-State Routing

In link-state routing, each router advertises the state and cost of its incident links. Reliable flooding distributes a link-state database, and every router independently computes shortest paths from the same topology view. The resulting forwarding tree is commonly generated using Dijkstra’s algorithm.

Link-state routing generally converges faster and supports richer topology-based metrics than basic distance-vector routing. Its costs are database memory, flooding overhead, computation and sensitivity to inconsistent or stale topology information. Hierarchical areas reduce the scale of flooding and shortest-path computation.

2.3 Path-Vector Routing

Path-vector routing advertises reachable prefixes together with an ordered sequence of autonomous systems or other policy-relevant attributes. BGP is the dominant example. A route is not selected solely by numerical shortest distance. Local preference, route origin, AS-path length, origin type, MED, eBGP/iBGP preference and router identifier may participate according to implementation and policy.

The AS path also provides loop prevention: an AS normally rejects a route advertisement containing its own AS number. However, BGP’s policy-driven behavior means that it is not a global shortest-path algorithm. A locally preferred route may be longer, more congested or less reliable than an alternative.

Comparison

Distance-vector protocols exchange reachability estimates; link-state protocols distribute topology information; path-vector protocols distribute paths and policy attributes. Link-state routing tends to optimize an internally consistent metric, whereas BGP intentionally supports administrative autonomy. No paradigm is universally superior: choice depends on scale, administrative boundaries, convergence requirements and policy complexity.

3. OSPF and IS-IS

3.1 OSPF

OSPF is a link-state interior gateway protocol. Routers exchange link-state advertisements (LSAs), construct a link-state database within an area, and run shortest-path computation. OSPF supports hierarchical operation through areas. Area 0 is the backbone through which non-backbone areas are connected in the standard hierarchical design.

Important OSPF concepts include:

• Router LSAs describing a router’s links within an area.
• Network LSAs generated for multi-access networks by a designated router.
• Area border routers connecting areas.
• Autonomous-system boundary routers redistributing or originating external routes.
• Equal-cost multipath, where multiple equal-cost next hops may be installed.

OSPF cost is an administrative metric. A common vendor convention derives cost from reference bandwidth divided by interface bandwidth, but the exact behavior is implementation-dependent and must be verified. The reference bandwidth should be configured consistently, particularly in high-speed networks; otherwise different interfaces may receive unintuitive or equal costs.

3.2 IS-IS

IS-IS is also a link-state protocol, originally developed for OSI networking and widely used for IP in service-provider networks. It uses a two-level hierarchy: Level 1 routers operate within an area, while Level 2 routers provide inter-area connectivity. A router may operate at both levels.

IS-IS carries information in type-length-value fields, making protocol extension comparatively flexible. It runs directly over the data-link layer rather than depending on IP for its own neighbor communication. Operationally, IS-IS and OSPF provide similar fundamental services, but their packet formats, area models, deployment traditions and extension mechanisms differ.

3.3 Design Considerations

OSPF or IS-IS design should consider area or level boundaries, summarization, failure domains, metric consistency, route redistribution and convergence behavior. Redistribution between protocols is a frequent source of feedback loops, suboptimal routing and loss of route information. Tagging, filtering and clear ownership of prefixes are essential.

4. BGP and Interdomain Policy

BGP establishes sessions over TCP and exchanges reachability information between autonomous systems. eBGP is used between autonomous systems, whereas iBGP distributes externally learned routes within an autonomous system. Full-mesh iBGP has poor scalability, motivating route reflectors and confederations.

A simplified route-selection sequence often considers highest local preference first, then shortest AS path, lowest origin type, lowest MED under applicable comparison rules, preference for eBGP over iBGP, and finally tie-breakers. Exact ordering is vendor- and configuration-sensitive; engineers must consult the implementation specification.

BGP policy supports:

• Customer, peer and provider relationships.
• Traffic engineering through local preference and route advertisements.
• Prefix filtering and maximum-prefix limits.
• Communities for scalable policy signaling.
• Route aggregation to reduce table size.
• Selective export and import control.

BGP’s flexibility creates risks. Route leaks can propagate unintended paths, and prefix hijacks can redirect traffic. Resource Public Key Infrastructure (RPKI) and Route Origin Authorizations help validate whether an AS is authorized to originate a prefix, but origin validation does not prove that the entire AS path is trustworthy. Complementary controls include prefix filters, monitoring, authenticated sessions where supported, and route-policy review.

5. Switching and Forwarding

Switching is the local forwarding function that maps an incoming packet or frame to an outgoing interface. Layer-2 Ethernet switches learn source MAC addresses, construct forwarding tables and flood unknown unicast, broadcast and relevant multicast traffic subject to VLAN rules. Layer-3 routers use destination prefixes and longest-prefix matching in a forwarding information base (FIB).

The control plane computes routes; the data plane applies forwarding decisions. A routing change is useful only after the resulting forwarding state is installed consistently enough to avoid blackholes, loops or transient microloops. Hardware forwarding may use TCAM or specialized lookup structures, while software forwarding offers greater flexibility but often lower packet-processing capacity.

VLANs separate Layer-2 broadcast domains. Spanning Tree Protocol prevents Layer-2 loops by blocking selected links, but it may leave capacity unused. Layer-3 fabrics, equal-cost multipath and overlay technologies are often used to improve utilization and scale. These choices introduce trade-offs involving encapsulation overhead, troubleshooting, control-plane complexity and failure isolation.

6. Multicast Routing

Unicast routing selects one path from one source to one destination. Multicast routing delivers traffic from a source to a set of receivers represented by a multicast group. Efficient multicast constructs a distribution tree rather than sending an independent unicast copy to each receiver.

A source-specific shortest-path tree is rooted at the source; a shared tree uses a rendezvous point or core. Protocol Independent Multicast (PIM) uses the unicast routing table for reverse-path forwarding checks. PIM-SM commonly begins with shared-tree behavior and may switch to source-specific trees when traffic justifies it. IGMP in IPv4 and MLD in IPv6 support receiver membership signaling on local networks.

Multicast optimization must consider bandwidth, state scalability, receiver dynamics and resilience. A tree minimizing total link cost may concentrate traffic on vulnerable links. Conversely, source-specific trees can improve path quality while consuming more state. Multicast also raises access-control and abuse concerns because unauthorized group membership can consume substantial capacity.

7. Traffic Engineering

Traffic engineering (TE) controls how traffic uses available resources to meet objectives such as utilization balance, latency bounds, resilience and service-level agreements. It may be distributed, centralized or hybrid.

A basic utilization measure for link \(e\) is

\[

\rho_e=\frac{f_e}{u_e},

\]

where \(f_e\) is carried flow and \(u_e\) is capacity, both measured in the same units, such as Mb/s. Values approaching one indicate limited headroom and increased sensitivity to bursts or failures.

Common TE mechanisms include MPLS traffic-engineered tunnels, RSVP-TE, segment routing, unequal policy choices in BGP, and centralized path computation. A useful objective may minimize the maximum utilization:

\[

\min U

\]

subject to

\[

f_e\leq Uu_e\quad \forall e.

\]

This objective differs from minimizing total cost. Minimum-cost routing can overload a few attractive links, while min-max utilization may distribute traffic over longer paths.

8. Constrained Shortest-Path Formulation

Let \(x_e\in\{0,1\}\) indicate whether directed edge \(e\) is selected. For a single route from source \(s\) to destination \(t\), valid path selection requires flow conservation:

\[

\sum_{e\in\delta^+(v)}x_e-

\sum_{e\in\delta^-(v)}x_e=

\begin{cases}

1,&v=s,\\

-1,&v=t,\\

0,&v\in V\setminus\{s,t\}.

\end{cases}

\]

Here \(\delta^+(v)\) and \(\delta^-(v)\) are outgoing and incoming edges. These constraints ensure one unit of flow leaves the source, enters the destination and is conserved at intermediate nodes. In a nonnegative-cost network, an optimal solution does not contain a positive-cost cycle. If zero-cost cycles are possible, additional simplicity constraints or a cycle-elimination argument may be needed when a simple path is explicitly required.

A constrained shortest-path model can be written as

\[

\min \sum_{e\in E}c_ex_e

\]

subject to flow conservation and constraints such as

\[

\sum_{e\in E}d_ex_e\leq D_{\max},

\]

\[

\sum_{e\in E}b_ex_e\leq B_{\max},

\]

\[

\sum_{e\in E}u^{\text{required}}_ex_e\geq U_{\min},

\]

where \(d_e\) is delay, \(b_e\) may represent administrative risk or resource consumption, and \(u^{\text{required}}_e\) is a suitability or reliability-related quantity only if its meaning is explicitly defined.

For independent link reliabilities, the correct reliability constraint is

\[

\prod_{e\in E}r_e^{x_e}\geq R_{\min}.

\]

Because \(x_e\) is binary, selected links contribute their reliability and unselected links contribute a factor of one. Taking logarithms gives the equivalent additive constraint

\[

\sum_{e\in E}[-\log r_e]x_e\leq -\log R_{\min}.

\]

This equivalence requires \(0<r_e\leq1\) and independent link failures under the path reliability model. Correlated failures require a richer model, such as shared-risk groups, scenario constraints or joint probability distributions. A simple sum \(\sum r_ex_e\geq R_{\min}\) is not a valid general path-reliability constraint.

The constrained shortest-path problem is generally harder than ordinary shortest path because multiple resource constraints can make the problem computationally difficult. Practical systems use label-setting or label-correcting methods, integer programming for selected demands, heuristics, constraint relaxation, or segment-routing candidate enumeration.

Worked Example 2: Feasibility with Reliability and Delay

Two paths connect \(s\) to \(t\):

• \(P_A\): cost 8, delay 12 ms, reliability \(0.98\times0.98=0.9604\).
• \(P_B\): cost 6, delay 18 ms, reliability \(0.995\times0.97=0.96515\).

Let the maximum delay be 20 ms and the minimum reliability be 0.96. Both paths are feasible. Since \(P_B\) has lower cost, it is selected by the stated objective, even though it has greater delay. If the reliability threshold were 0.97, neither path would be feasible. This illustrates why feasibility must be checked before optimization.

9. Segment Routing

Segment routing encodes a path as an ordered list of instructions called segments. In an MPLS data plane, segments may be represented by labels; in IPv6, Segment Routing over IPv6 (SRv6) uses a segment list in an IPv6 extension mechanism. A node segment generally identifies a shortest path to a node, while an adjacency segment can force traversal of a particular link or adjacency.

The ingress node imposes the segment list, reducing dependence on per-tunnel signaling and potentially simplifying state in transit routers. An optimization controller can compute a path satisfying latency, disjointness or resource constraints and then encode it as a segment list. However, segment lists can become long, node-segment paths may not express every desired constraint, and failures require valid local repair or recomputation.

Fast reroute techniques, including local protection concepts, aim to restore forwarding rapidly after failure. Protection may be link-disjoint, node-disjoint or shared-risk-disjoint. These terms are not interchangeable: two paths can be link-disjoint but share a conduit or power system.

10. Optimization Objectives and Multi-Criteria Decisions

Network optimization rarely has one universally correct objective. Candidate objectives include:

• Minimum administrative cost.
• Minimum delay or jitter.
• Maximum residual capacity.
• Minimum maximum link utilization.
• Maximum reliability or availability.
• Minimum energy consumption.
• Policy compliance and security exposure.

A weighted objective such as

\[

\min \sum_e(\alpha c_e+\beta d_e+\gamma[-\log r_e])x_e

\]

is meaningful only when the terms are normalized or assigned compatible units. Otherwise, a coefficient may merely compensate for arbitrary measurement scales. An alternative is lexicographic optimization: first require policy and feasibility, then minimize risk, then optimize cost, and finally use delay as a tie-breaker.

Pareto analysis is useful when no single solution dominates all others. A path is Pareto-optimal if no other feasible path is at least as good in every criterion and strictly better in one. The final choice may depend on service-level agreements, operational simplicity and the cost of failure.

11. Illustrative Case: Interregional Service-Provider Upgrade

Background

Consider a fictional service provider connecting three metropolitan regions through a core network. Existing interior routing selects minimum-cost paths, but evening traffic creates 82% utilization on one core link while parallel links remain below 35%. A latency-sensitive enterprise service also requires no more than 25 ms one-way delay and a path reliability of at least 0.97 under an independent-link planning model.

Problem

The provider must improve utilization without violating latency, reliability or administrative policy. Directly changing all IGP metrics could affect thousands of ordinary flows and create unpredictable changes.

Analysis

The engineering team builds a graph with link capacities, measured delays, administrative costs and planned reliabilities. Candidate paths are enumerated. For each path, the team checks flow conservation implicitly through path construction, computes delay as the sum of link delays, and computes reliability as the product of link reliabilities. Paths sharing a conduit are marked as sharing a risk group and are not treated as fully diverse.

The least-cost path has 18 ms delay and reliability 0.975 but would increase the already congested link to 91% utilization. A second path has 22 ms delay and reliability 0.978 and uses a parallel core corridor. A third path has 28 ms delay and is rejected by the latency constraint.

Approach and Solution

The provider deploys segment routing. The second path is encoded using node and adjacency segments, with a backup segment list selected for local repair. A controller monitors utilization and recomputes candidate paths when demand forecasts indicate that utilization may exceed 75%. The ordinary IGP remains unchanged, limiting unintended impact on other traffic.

Outcome

Peak utilization on the congested corridor falls from 82% to 68%, while the enterprise path remains within 22 ms delay and 0.978 reliability under the planning model. The provider documents that the reliability value is conditional on the independence assumption and separately records shared-risk information for maintenance planning.

Learning Points

The case demonstrates that shortest path, reliability, capacity and policy are distinct criteria. Segment routing can express a selected path without requiring every transit router to maintain a separate tunnel state, but it does not remove the need for accurate topology, capacity and failure-domain data.

12. Scalability, Operations and Governance

Routing scalability depends on topology size, routing-table growth, update frequency, flooding scope, path-computation complexity and forwarding-table capacity. Hierarchical areas, route summarization, route reflectors, carefully bounded policy and aggregation reduce scale, but summarization can hide topology detail and create suboptimal paths.

Convergence engineering must consider detection time, computation time, control-plane propagation, FIB installation and transient forwarding behavior. Fast timers may detect failures sooner but increase false positives and control traffic. Monitoring should include loss, delay, jitter, utilization, route changes, control-plane CPU, adjacency state and policy violations.

Security controls include authenticated routing sessions where supported, infrastructure ACLs, TTL security mechanisms, prefix filtering, RPKI-based origin validation, control-plane policing and change management. Routing policy is also a governance issue: automated optimization must be auditable, reversible and bounded by approved constraints. A controller should not optimize solely for utilization while violating contractual isolation, data residency, emergency-service priority or security policy.

Artificial intelligence can assist demand forecasting, anomaly detection and candidate-path ranking. It should not be treated as an unquestionable routing authority. Training data may encode transient failures or biased operational decisions; model outputs require constraint verification, explainability, rollback and human oversight. A deterministic feasibility checker should remain in the control loop for safety-critical constraints.

13. Best-Practice Design Principles

1. Separate reachability, policy and optimization objectives.

2. Use consistent metric definitions and verify units.

3. Include complete flow-conservation constraints in path formulations.

4. Model reliability multiplicatively or through a justified logarithmic transformation.

5. Represent shared-risk failures explicitly rather than assuming independent links.

6. Design hierarchy and summarization with known information-loss effects.

7. Validate BGP import and export policy in both directions.

8. Maintain headroom for bursts, failures and maintenance.

9. Test convergence and microloop behavior before production deployment.

10. Make automated changes observable, auditable and reversible.

Challenges and Limitations

No routing protocol simultaneously provides perfect global optimality, instantaneous convergence, minimal state and complete policy flexibility. Distributed protocols operate with incomplete or delayed information. Centralized optimization can improve global coordination but introduces controller reachability, telemetry quality, computational and trust dependencies. Reliability estimates can be inaccurate, especially when common-cause failures dominate. Multi-constraint routing can be computationally difficult, and operational objectives may conflict. Finally, route optimization cannot compensate for inadequate capacity, poor physical diversity or incorrect topology data.

Chapter Summary

Routing protocols exchange different forms of information: distance-vector protocols exchange distance estimates, link-state protocols distribute topology, and path-vector protocols distribute policy-bearing paths. OSPF and IS-IS provide link-state interior routing, while BGP provides policy-driven interdomain routing. Switching transforms control-plane decisions into local data-plane forwarding, and multicast constructs distribution trees for receiver groups. Traffic engineering selects paths to balance capacity, delay, reliability and policy rather than merely minimizing a default metric.

Graph theory provides a rigorous basis for routing. A valid binary path formulation requires source–destination flow conservation. Additive metrics can be summed directly, but independent-link reliability is multiplicative and must be modeled as a product or transformed with negative logarithms. Segment routing allows a computed path to be represented as a compact sequence of forwarding instructions. Effective network optimization combines mathematical feasibility, protocol behavior, scalability, security, operational monitoring and governance.

Key Terms

Graph: Mathematical representation of nodes and links.

Edge: A directed or undirected graph link.

Flow conservation: Constraint ensuring that selected flow enters and leaves intermediate nodes consistently.

Distance-vector routing: Routing based on neighbor exchanges of destination-distance estimates.

Link-state routing: Routing based on flooded topology information and local shortest-path computation.

Path-vector routing: Routing based on advertised paths and policy attributes.

OSPF: Link-state interior gateway protocol for IP networks.

IS-IS: Two-level link-state routing protocol widely used in provider networks.

BGP: Policy-driven path-vector protocol for interdomain routing.

FIB: Forwarding information base used by the data plane.

Multicast: Delivery from one source to multiple receivers through a distribution structure.

Traffic engineering: Deliberate control of traffic paths and resource use.

Segment routing: Source-influenced forwarding using an ordered segment list.

Equal-cost multipath: Forwarding over multiple paths with equal routing cost.

Reliability: Probability that a service or path performs its required function.

Shared-risk link group: Set of links exposed to a common failure cause.

Route reflector: BGP mechanism that reduces the need for full-mesh iBGP.

Constrained shortest path: Path optimization subject to one or more resource constraints.

Pareto-optimal path: Feasible path not dominated across all considered objectives.

Microloop: Temporary forwarding loop during a routing convergence event.

Review Questions

Short Answer Questions

1. Distinguish distance-vector, link-state and path-vector routing.

2. What is the purpose of flow conservation in a path optimization model?

3. Why is path reliability generally multiplicative rather than additive?

4. State two scalability mechanisms used in BGP.

5. What is the role of a segment list in segment routing?

Long Answer Questions

1. Explain OSPF and IS-IS operation, hierarchy and design trade-offs.

2. Discuss BGP route selection, policy control and security limitations.

3. Compare shortest-path routing with min-max utilization traffic engineering.

4. Explain multicast tree construction and its scalability challenges.

5. Evaluate the benefits and limitations of centralized path computation.

Analytical / Application Questions

1. Formulate a binary constrained path model with cost, delay, capacity and independent-link reliability constraints. Explain every constraint and assumption.

2. A network has two feasible paths: one minimizes cost and the other minimizes maximum utilization. Specify additional information needed to select between them and justify a decision framework.

3. Design a segment-routing-based recovery strategy for a service whose primary and backup paths must be node-disjoint and must avoid a shared-risk physical corridor.

Multiple Choice Questions with Answers

Question 1

Which information is primarily distributed by a link-state routing protocol?

A. Only the next-hop distance to each destination

B. The topology and state of local links

C. Only the AS path of external prefixes

D. The complete packet payload

Correct Answer: B. The topology and state of local links

Question 2

For independent link reliabilities, how is the reliability of a path calculated?

A. By summing link reliabilities

B. By multiplying link reliabilities

C. By taking the largest link reliability

D. By averaging link reliabilities

Correct Answer: B. By multiplying link reliabilities

Question 3

Which constraint ensures that a binary edge-selection solution represents an s-to-t flow?

A. A source–destination flow-conservation constraint

B. A route-reflector constraint

C. A VLAN tagging constraint

D. A maximum-prefix constraint

Correct Answer: A. A source–destination flow-conservation constraint

Question 4

What is the principal role of BGP?

A. Intra-switch MAC learning

B. Interdomain, policy-driven route exchange

C. Local multicast membership signaling

D. Physical-layer error correction

Correct Answer: B. Interdomain, policy-driven route exchange

Question 5

If two links have reliabilities 0.98 and 0.97 and failures are independent, their path reliability is:

A. 0.9506

B. 0.9750

C. 1.9500

D. 0.0194

Correct Answer: A. 0.9506

Question 6

Which objective directly minimizes the worst link utilization?

A. Minimize the maximum value of f_e/u_e

B. Minimize the number of routing protocols

C. Maximize the AS-path length

D. Minimize the number of destination prefixes

Correct Answer: A. Minimize the maximum value of f_e/u_e

Question 7

What does a segment-routing adjacency segment generally do?

A. Forces traversal of a specified adjacency

B. Removes all forwarding state from the ingress router

C. Advertises a BGP community

D. Creates an Ethernet broadcast domain

Correct Answer: A. Forces traversal of a specified adjacency

Question 8

Which statement about a simple sum of link reliabilities is correct?

A. It is always the exact path reliability

B. It is generally not a valid path-reliability model

C. It is required by OSPF

D. It applies only to multicast trees

Correct Answer: B. It is generally not a valid path-reliability model

Question 9

What is a major purpose of BGP route reflectors?

A. To eliminate IP addressing

B. To reduce the full-mesh requirement of iBGP

C. To compute Ethernet spanning trees

D. To measure optical power

Correct Answer: B. To reduce the full-mesh requirement of iBGP

Question 10

Why may minimum-cost routing fail to minimize congestion?

A. It can concentrate flows on attractive low-cost links

B. It always selects paths with the most hops

C. It ignores destination addresses

D. It cannot use link-state information

Correct Answer: A. It can concentrate flows on attractive low-cost links