Question 670A hospital uses a machine-learning capacity predictor without preserving the features that generated each recommendation. The design team recommends: Store controlled feature provenance and validation evidence alongside approved capacity decisions. Which architectural principle best supports that recommendation? AI-assisted architecture must be explainable enough to audit design-driving predictions. Statistical patterns learned from one network regime may not generalize to rare operating conditions. Dual-stack availability must be engineered, not presumed from IPv4-only success. AI inference scaling requires corresponding network endpoint lifecycle and reachability automation. ExplanationCorrect Answer A:AI-assisted architecture must be explainable enough to audit design-driving predictions. Store controlled feature provenance and validation evidence alongside approved capacity decisions. That is the direct response to this rule. Verify using: A reviewed expansion proposal identifies its training inputs, assumptions and error bounds. Incorrect Answers B:Statistical patterns learned from one network regime may not generalize to rare operating conditions. This does not establish the proposed intervention. The governing principle here is: AI-assisted architecture must be explainable enough to audit design-driving predictions. Apply it by: Store controlled feature provenance and validation evidence alongside approved capacity decisions. C:Dual-stack availability must be engineered, not presumed from IPv4-only success. This does not establish the proposed intervention. The governing principle here is: AI-assisted architecture must be explainable enough to audit design-driving predictions. Apply it by: Store controlled feature provenance and validation evidence alongside approved capacity decisions. D:AI inference scaling requires corresponding network endpoint lifecycle and reachability automation. This does not establish the proposed intervention. The governing principle here is: AI-assisted architecture must be explainable enough to audit design-driving predictions. Apply it by: Store controlled feature provenance and validation evidence alongside approved capacity decisions.
Question 671A hospital merger links three overlapping OSPF domains and proposes redistributing every external prefix at every boundary. After the proposed design change — Limit redistribution to approved routes with tagging and loop prevention at defined transition points. — which outcome would provide the strongest validation? A failover exercise confirms one valid route origin and no persistent loop for clinical prefixes. A controlled exit shutdown moves representative commerce flows without creating asymmetric firewall state. Pilot measurements show acceptable query latency and documented per-request transfer costs. A full-site outage drill brings up transactions with independent DNS and trust infrastructure. ExplanationCorrect Answer A:A failover exercise confirms one valid route origin and no persistent loop for clinical prefixes. The evidence matches the stated repair. Uncontrolled mutual redistribution creates reachability ambiguity and can circulate stale routes. Incorrect Answers B:External route resilience requires predictable return paths and convergence behavior through each surviving exit. This relates to a separate check and does not verify the specified repair. The relevant acceptance evidence is: A failover exercise confirms one valid route origin and no persistent loop for clinical prefixes. C:Pilot measurements show acceptable query latency and documented per-request transfer costs. This is not a test of the affected control. Uncontrolled mutual redistribution creates reachability ambiguity and can circulate stale routes. The required check is: A failover exercise confirms one valid route origin and no persistent loop for clinical prefixes. D:Application recovery is incomplete when its control services share the original failure domain. This relates to a separate check and does not verify the specified repair. The relevant acceptance evidence is: A failover exercise confirms one valid route origin and no persistent loop for clinical prefixes.
Question 672A government agency evaluates hosted and on-prem language models solely on initial infrastructure price. After the proposed design change — Compare lifecycle compute, network, data-governance and audit obligations across placement options. — which outcome would provide the strongest validation? A test out-of-policy recommendation is rejected and the prior forwarding state can be restored. The decision record compares whole-life emissions and identifies equipment with justified early retirement. The decision compares per-request cost and compliance outcomes at representative demand levels. A combined design review resolves common IP space and identity flows before either team deploys. ExplanationCorrect Answer C:The decision compares per-request cost and compliance outcomes at representative demand levels. This tests the affected component rather than an unrelated service. AI deployment economics require ROI and operational exposure modeling beyond acquisition cost. Without the change, a superficially cheap platform could create unbudgeted governance and egress expense. Incorrect Answers A:AI-driven optimization requires deterministic guardrails for high-impact network changes. This relates to a separate check and does not verify the specified repair. The relevant acceptance evidence is: The decision compares per-request cost and compliance outcomes at representative demand levels. B:A sustainable refresh balances operational energy with manufacturing impact and continued service requirements. This relates to a separate check and does not verify the specified repair. The relevant acceptance evidence is: The decision compares per-request cost and compliance outcomes at representative demand levels. D:Parallel design work needs stable dependency contracts even when components evolve independently. This relates to a separate check and does not verify the specified repair. The relevant acceptance evidence is: The decision compares per-request cost and compliance outcomes at representative demand levels.
Question 673A manufacturer builds VXLAN overlays but sizes underlay links without encapsulation and replication overhead. Which network architecture decision best meets these requirements? Document authentication cache and break-glass behavior for isolated network management segments. Define single-writer ownership and conflict handling for each network intent object. Apply tiered retention, sampling and critical-metric preservation tied to diagnostic use cases. Budget MTU and bandwidth for overlay encapsulation and broadcast/unknown handling. ExplanationCorrect Answer D:Budget MTU and bandwidth for overlay encapsulation and broadcast/unknown handling. This targets the stated fault because overlay network performance inherits underlay constraints plus extra protocol overhead. After repair, representative encapsulated packets avoid MTU drops and leaf uplinks stay below peak saturation. Incorrect Answers A:Management-plane dependency resilience includes identity services and auditable emergency controls. That is relevant to a different intervention. For the stated situation, the required action is: Budget MTU and bandwidth for overlay encapsulation and broadcast/unknown handling. The governing reason is: Overlay network performance inherits underlay constraints plus extra protocol overhead. B:Multiple orchestration systems require explicit source-of-truth and reconciliation boundaries. That is relevant to a different intervention. For the stated situation, the required action is: Budget MTU and bandwidth for overlay encapsulation and broadcast/unknown handling. The governing reason is: Overlay network performance inherits underlay constraints plus extra protocol overhead. C:Observability architecture must balance useful fidelity with ingestion and storage sustainability. That is relevant to a different intervention. For the stated situation, the required action is: Budget MTU and bandwidth for overlay encapsulation and broadcast/unknown handling. The governing reason is: Overlay network performance inherits underlay constraints plus extra protocol overhead.
Question 674A vector search application adds GPU inference instances dynamically but routing uses long-lived static backend entries. The design team recommends: Provide dynamic service discovery and health-aware load distribution for scaled inference endpoints. Which architectural principle best supports that recommendation? AI inference scaling requires corresponding network endpoint lifecycle and reachability automation. Model input secrecy must persist through network observability and support systems. Provider branding does not guarantee physical-route diversity. Generated design advice is unreliable without accurate trusted network context. ExplanationCorrect Answer A:AI inference scaling requires corresponding network endpoint lifecycle and reachability automation. This links the root condition to the proposed change. Provide dynamic service discovery and health-aware load distribution for scaled inference endpoints. The expected result is: A synthetic scale-out adds healthy inference backends without manual route edits or stale responses. Incorrect Answers B:Model input secrecy must persist through network observability and support systems. This does not establish the proposed intervention. The governing principle here is: AI inference scaling requires corresponding network endpoint lifecycle and reachability automation. Apply it by: Provide dynamic service discovery and health-aware load distribution for scaled inference endpoints. C:Provider branding does not guarantee physical-route diversity. This does not establish the proposed intervention. The governing principle here is: AI inference scaling requires corresponding network endpoint lifecycle and reachability automation. Apply it by: Provide dynamic service discovery and health-aware load distribution for scaled inference endpoints. D:Generated design advice is unreliable without accurate trusted network context. This does not establish the proposed intervention. The governing principle here is: AI inference scaling requires corresponding network endpoint lifecycle and reachability automation. Apply it by: Provide dynamic service discovery and health-aware load distribution for scaled inference endpoints.
Question 675A logistics enterprise requests an AI-generated network topology optimization without disclosing that one route carries safety-critical traffic. After the proposed design change — Encode immovable safety constraints and test proposed optimization against business traffic classification. — which outcome would provide the strongest validation? A full-site outage drill brings up transactions with independent DNS and trust infrastructure. A simultaneous checkpoint rollout stays inside throughput caps while online queries meet latency goals. A generated shortcut that violates safety path isolation is rejected before deployment. Operations staff can diagnose a controlled EVPN neighbor loss using the delivered runbooks. ExplanationCorrect Answer C:A generated shortcut that violates safety path isolation is rejected before deployment. Automated design suggestions must operate inside explicitly documented non-negotiable constraints. The test establishes that the fix addresses the specific failure. Incorrect Answers A:Application recovery is incomplete when its control services share the original failure domain. This relates to a separate check and does not verify the specified repair. The relevant acceptance evidence is: A generated shortcut that violates safety path isolation is rejected before deployment. B:Model checkpoint distribution can generate bandwidth bursts far larger than inference requests. This relates to a separate check and does not verify the specified repair. The relevant acceptance evidence is: A generated shortcut that violates safety path isolation is rejected before deployment. D:Operations staff can diagnose a controlled EVPN neighbor loss using the delivered runbooks. An unrelated service might produce that result while the failure continues. The relevant proof is: A generated shortcut that violates safety path isolation is rejected before deployment.
Question 676A retail company plans global cloud DR but does not reconcile its operational recovery-point target with database replication latency. After the proposed design change — Choose a recovery topology and data movement mode matched to the approved RPO and RTO. — which outcome would provide the strongest validation? Loss of the primary transit service leaves ticketing application calls within recovery limits. A peak replication trial meets committed-write timing and recovery point objectives. A representative deployment meets latency objectives with documented transfer costs per workload. A planned failover measures accepted data age and service restoration within the business tolerances. ExplanationCorrect Answer D:A planned failover measures accepted data age and service restoration within the business tolerances. This is the required observable result. Choose a recovery topology and data movement mode matched to the approved RPO and RTO. The original risk was: A remote copy could look available yet contain unacceptable transaction gaps. Incorrect Answers A:Cloud-native component redundancy is insufficient when all paths share a single network chokepoint. This relates to a separate check and does not verify the specified repair. The relevant acceptance evidence is: A planned failover measures accepted data age and service restoration within the business tolerances. B:A peak replication trial meets committed-write timing and recovery point objectives. This is not a test of the affected control. Cloud disaster recovery choices need measurable consistency and failover behavior, not just multiple regions. The required check is: A planned failover measures accepted data age and service restoration within the business tolerances. C:Cloud network topology and application placement affect recurring usage charges. This relates to a separate check and does not verify the specified repair. The relevant acceptance evidence is: A planned failover measures accepted data age and service restoration within the business tolerances.
Question 677A logistics provider asks an architect to select a router model before collecting any growth, path or service objectives. After the proposed design change — Conduct a structured discovery of traffic, sites, availability and operational constraints before selecting platforms. — which outcome would provide the strongest validation? The team signs stable underlay interfaces and reviews each new store requirement against an agreed change budget. Peak-load tests complete fraud scoring inside the transaction budget and exercise the fallback mode. A reviewed requirements register supports sizing calculations and at least two viable technology options. Carrier evidence identifies distinct entrance facilities and a controlled failure test maintains control traffic. ExplanationCorrect Answer C:A reviewed requirements register supports sizing calculations and at least two viable technology options. The evidence matches the stated repair. Requirement-driven design starts with outcomes and constraints rather than product selection. Incorrect Answers A:A hybrid governance cadence can preserve safety-critical interfaces while admitting bounded requirement changes. This relates to a separate check and does not verify the specified repair. The relevant acceptance evidence is: A reviewed requirements register supports sizing calculations and at least two viable technology options. B:Peak-load tests complete fraud scoring inside the transaction budget and exercise the fallback mode. An unrelated service might produce that result while the failure continues. The relevant proof is: A reviewed requirements register supports sizing calculations and at least two viable technology options. D:Logical path redundancy does not remove shared physical failure modes. This relates to a separate check and does not verify the specified repair. The relevant acceptance evidence is: A reviewed requirements register supports sizing calculations and at least two viable technology options.
Question 678A university uses shared wireless guest credentials and expects to identify individual users during security investigations. Which network architecture decision best meets these requirements? Select fail-open or fail-closed behavior by service risk and design a resilient, testable enforcement path. Distribute critical policy gateways across independent facility failure domains and confirm consistent rules. Use accountable guest identity or properly scoped anonymous access with separate limited privileges. Use device compliance posture and identity-aware segmentation with a separate BYOD access tier. ExplanationCorrect Answer C:Use accountable guest identity or properly scoped anonymous access with separate limited privileges. Authentication and attribution goals must match the actual business and privacy requirements. The practical confirmation is: An incident record identifies an authorized guest session without exposing unrelated personal data. Incorrect Answers A:Inline security services affect availability and confidentiality through their failure mode. That is relevant to a different intervention. For the stated situation, the required action is: Use accountable guest identity or properly scoped anonymous access with separate limited privileges. The governing reason is: Authentication and attribution goals must match the actual business and privacy requirements. B:Security service availability depends on power and site topology as well as policy logic. That is relevant to a different intervention. For the stated situation, the required action is: Use accountable guest identity or properly scoped anonymous access with separate limited privileges. The governing reason is: Authentication and attribution goals must match the actual business and privacy requirements. D:User authorization does not imply device trust, especially on sensitive shared research networks. That is relevant to a different intervention. For the stated situation, the required action is: Use accountable guest identity or properly scoped anonymous access with separate limited privileges. The governing reason is: Authentication and attribution goals must match the actual business and privacy requirements.
Question 679A financial trading platform asks for subsecond failover yet proposes routing convergence based only on default protocol dead timers. After the proposed design change — Use supported rapid fault detection and precomputed alternate paths appropriate to the service SLA. — which outcome would provide the strongest validation? A stale resolver is withdrawn and clients receive consistent authorized records from healthy sites. A controlled trace retains performance metadata without exposing protected prompt contents. A simulated application failure withdraws or redirects the service while another site remains functional. A link failure meets measured restoration timing without instability on normal transients. ExplanationCorrect Answer D:A link failure meets measured restoration timing without instability on normal transients. The observation confirms the corrected dependency: Recovery time is the sum of fault detection, route computation and forwarding installation. The avoided exposure is: Slow convergence could exceed market transaction outage limits. Incorrect Answers A:Anycast reachability is safe only when serving sites have sufficiently consistent data and health checks. This relates to a separate check and does not verify the specified repair. The relevant acceptance evidence is: A link failure meets measured restoration timing without instability on normal transients. B:Model input secrecy must persist through network observability and support systems. This relates to a separate check and does not verify the specified repair. The relevant acceptance evidence is: A link failure meets measured restoration timing without instability on normal transients. C:Network reachability does not prove that the destination application can serve real transactions. This relates to a separate check and does not verify the specified repair. The relevant acceptance evidence is: A link failure meets measured restoration timing without instability on normal transients.
Question 680An engineering firm evaluates IaaS lift-and-shift without accounting for traffic charges between cloud availability zones. The design team recommends: Include east-west and inter-zone egress patterns in placement and economic scenarios. Which architectural principle best supports that recommendation? Cloud-hosted network services need separate business and management security boundaries. Application health metrics must include both signaling and the media they establish. Cloud network topology and application placement affect recurring usage charges. Hybrid routing should expose the minimum necessary address space across trust boundaries. ExplanationCorrect Answer C:Cloud network topology and application placement affect recurring usage charges. This links the root condition to the proposed change. Include east-west and inter-zone egress patterns in placement and economic scenarios. The expected result is: A representative deployment meets latency objectives with documented transfer costs per workload. Incorrect Answers A:Cloud-hosted network services need separate business and management security boundaries. This does not establish the proposed intervention. The governing principle here is: Cloud network topology and application placement affect recurring usage charges. Apply it by: Include east-west and inter-zone egress patterns in placement and economic scenarios. B:Application health metrics must include both signaling and the media they establish. This does not establish the proposed intervention. The governing principle here is: Cloud network topology and application placement affect recurring usage charges. Apply it by: Include east-west and inter-zone egress patterns in placement and economic scenarios. D:Hybrid routing should expose the minimum necessary address space across trust boundaries. This does not establish the proposed intervention. The governing principle here is: Cloud network topology and application placement affect recurring usage charges. Apply it by: Include east-west and inter-zone egress patterns in placement and economic scenarios.