cert
cert-1
cert-2

Pass Cisco 300-640 Exam in First Attempt Guaranteed!

Get 100% Latest Exam Questions, Accurate & Verified Answers to Pass the Actual Exam!
30 Days Free Updates, Instant Download!

cert-5
cert-6
300-640 Exam - Verified By Experts
300-640 Premium File

300-640 Premium File

$89.99
$98.99
  • Premium File 60 Questions & Answers. Last Update: Nov 10, 2026

Whats Included:

  • Latest Questions
  • 100% Accurate Answers
  • Fast Exam Updates
 
$98.99
$89.99
accept 10 downloads in the last 7 days
block-screenshots
Cisco 300-640 Practice Test 20

Question 190

The acceptance board must decide whether an AI-infrastructure response to gpu failure-domain placement is fit for controlled operation. The fictional operating facts are: A municipal analytics platform spreads GPU pods across hosts but stores all checkpoints on one unprotected local disk. Signoff cannot rely on a good-looking utilization graph, a change ticket alone or an unmeasured promise. The response must include an appropriate governing engineering principle, a scoped intervention and a separate measurable acceptance criterion. Choose EXACTLY THREE statements that meet those roles.

  1. A planned node or rack loss restarts the job from an approved checkpoint within recovery objectives.
  2. The same-load measurement demonstrates reduced facility overhead without harming service performance.
  3. Use workload-aware anti-affinity and recovery checkpoints across independent domains where supported.
  4. Distributed training requires adequate effective east-west throughput between GPU workers, with oversubscription based on real collective traffic.
  5. Distributed AI placement must limit correlated failures and balance locality against rack, host and power dependencies.
  6. Build physically independent eligible paths and test controlled link or leaf failure under active AI traffic.

Explanation

Correct Answer

A:

A municipal analytics platform spreads GPU pods across hosts but stores all checkpoints on one unprotected local disk. A planned node or rack loss restarts the job from an approved checkpoint within recovery objectives. Use workload-aware anti-affinity and recovery checkpoints across independent domains where supported.

C:

A municipal analytics platform spreads GPU pods across hosts but stores all checkpoints on one unprotected local disk. Use workload-aware anti-affinity and recovery checkpoints across independent domains where supported. Distributed AI placement must limit correlated failures and balance locality against rack, host and power dependencies.

E:

A municipal analytics platform spreads GPU pods across hosts but stores all checkpoints on one unprotected local disk. Distributed AI placement must limit correlated failures and balance locality against rack, host and power dependencies. Use workload-aware anti-affinity and recovery checkpoints across independent domains where supported.

Incorrect Answers

B:

The same-load measurement demonstrates reduced facility overhead without harming service performance. It would not independently demonstrate that the control works: A municipal analytics platform spreads GPU pods across hosts but stores all checkpoints on one unprotected local disk. A planned node or rack loss restarts the job from an approved checkpoint within recovery objectives. Use workload-aware anti-affinity and recovery checkpoints across independent domains where supported.

D:

Distributed training requires adequate effective east-west throughput between GPU workers, with oversubscription based on real collective traffic. It is not the rule that supports the proposed change and acceptance test: A municipal analytics platform spreads GPU pods across hosts but stores all checkpoints on one unprotected local disk. Distributed AI placement must limit correlated failures and balance locality against rack, host and power dependencies. Use workload-aware anti-affinity and recovery checkpoints across independent domains where supported.

F:

Build physically independent eligible paths and test controlled link or leaf failure under active AI traffic. It does not implement the governing rule for this approval: A municipal analytics platform spreads GPU pods across hosts but stores all checkpoints on one unprotected local disk. Use workload-aware anti-affinity and recovery checkpoints across independent domains where supported. A planned node or rack loss restarts the job from an approved checkpoint within recovery objectives.

Question 191

A data-center AI architecture committee considers accelerator virtualization support in its AI Infrastructure Components and Architecture review. At the affected site: A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. The committee is deliberately not selecting a vendor product or applying a live change yet. It needs the foundational reasoning that makes the next design decision defensible. Which ONE engineering principle directly governs this situation? Choose ONE answer.

  1. GPU partitioning and virtualization require platform-licensed, driver-compatible capabilities verified for the target workload.
  2. A low-average-latency network can still fail serving SLOs when transient queuing creates high p99 latency.
  3. AI servers, firmware, NIC drivers, GPU stack and fabric software form a versioned support matrix.
  4. Distributed training requires adequate effective east-west throughput between GPU workers, with oversubscription based on real collective traffic.

Explanation

Correct Answer

A:

A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. GPU partitioning and virtualization require platform-licensed, driver-compatible capabilities verified for the target workload. Check firmware, drivers, hypervisor or container runtime compatibility, partition inventory and isolation tests.

Incorrect Answers

B:

A low-average-latency network can still fail serving SLOs when transient queuing creates high p99 latency. It does not govern the condition described: A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. GPU partitioning and virtualization require platform-licensed, driver-compatible capabilities verified for the target workload.

C:

AI servers, firmware, NIC drivers, GPU stack and fabric software form a versioned support matrix. It does not govern the condition described: A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. GPU partitioning and virtualization require platform-licensed, driver-compatible capabilities verified for the target workload.

D:

Distributed training requires adequate effective east-west throughput between GPU workers, with oversubscription based on real collective traffic. It does not govern the condition described: A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. GPU partitioning and virtualization require platform-licensed, driver-compatible capabilities verified for the target workload.

Question 192

An infrastructure diagnostician is investigating accelerator virtualization support for the AI Infrastructure Components and Architecture team. The reported incident reads: A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. Competing teams attribute the symptoms to different layers of the AI stack. Before accepting a hypothesis or touching production settings, the investigator must choose a discriminating set of measurements. Which ONE evidence package is MOST relevant? Choose ONE answer.

  1. Map node-to-rack, PDU, network-fabric and storage failure domains alongside worker-placement requirements.
  2. Record adapter firmware, host kernel, accelerator driver, CUDA/runtime level, Cisco UCS profiles and network release compatibility.
  3. Compare dataset manifests, content hashes, model versions, replication lag and divergent write histories.
  4. Check firmware, drivers, hypervisor or container runtime compatibility, partition inventory and isolation tests.

Explanation

Correct Answer

D:

A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. Check firmware, drivers, hypervisor or container runtime compatibility, partition inventory and isolation tests. GPU partitioning and virtualization require platform-licensed, driver-compatible capabilities verified for the target workload.

Incorrect Answers

A:

Map node-to-rack, PDU, network-fabric and storage failure domains alongside worker-placement requirements. That check would not establish the decisive facts for this condition: A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. Check firmware, drivers, hypervisor or container runtime compatibility, partition inventory and isolation tests.

B:

Record adapter firmware, host kernel, accelerator driver, CUDA/runtime level, Cisco UCS profiles and network release compatibility. That check would not establish the decisive facts for this condition: A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. Check firmware, drivers, hypervisor or container runtime compatibility, partition inventory and isolation tests.

C:

Compare dataset manifests, content hashes, model versions, replication lag and divergent write histories. That check would not establish the decisive facts for this condition: A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. Check firmware, drivers, hypervisor or container runtime compatibility, partition inventory and isolation tests.

Question 193

During a bounded AI-platform change-control review concerning accelerator virtualization support, the site owner reports: A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. The proposed work must have a narrow implementation scope, first be validated against the failure evidence and avoid unrelated system-wide tuning. Which ONE intervention should the engineers pilot before any wider rollout? Choose ONE answer.

  1. Set class-based retention and quotas while keeping legally required provenance and restorable checkpoints.
  2. Validate and stage a supported end-to-end software bill of materials with rollback images.
  3. Revise supported precision, checkpointing or model sharding and select GPU memory capacity based on peak use.
  4. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior.

Explanation

Correct Answer

D:

A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior. Check firmware, drivers, hypervisor or container runtime compatibility, partition inventory and isolation tests.

Incorrect Answers

A:

Set class-based retention and quotas while keeping legally required provenance and restorable checkpoints. That change would leave the reported problem unresolved: A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior.

B:

Validate and stage a supported end-to-end software bill of materials with rollback images. That change would leave the reported problem unresolved: A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior.

C:

Revise supported precision, checkpointing or model sharding and select GPU memory capacity based on peak use. That change would leave the reported problem unresolved: A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior.

Question 194

Two independent technical reviewers examine a AI Infrastructure Components and Architecture problem involving accelerator virtualization support. Their shared incident log states: A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. One reviewer requires measurements capable of separating the observed problem from neighboring failure modes; the other requires a bounded action that follows those measurements. Choose EXACTLY TWO statements that together supply those two parts.

  1. Use versioned immutable datasets with controlled transfer promotion and reject stale job inputs.
  2. Map VLAN or VRF boundaries, workload identities, firewall rules, east-west flows and attempted cross-tenant connections.
  3. Check firmware, drivers, hypervisor or container runtime compatibility, partition inventory and isolation tests.
  4. Observe GPU idle gaps, CPU saturation, page faults, dataloader queue depth, NUMA affinity and PCIe locality.
  5. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior.

Explanation

Correct Answer

C:

A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. Check firmware, drivers, hypervisor or container runtime compatibility, partition inventory and isolation tests. Multiple tenants operate within their allocations while a full-device benchmark remains unaffected.

E:

A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior. Multiple tenants operate within their allocations while a full-device benchmark remains unaffected.

Incorrect Answers

A:

Use versioned immutable datasets with controlled transfer promotion and reject stale job inputs. It is not the scoped response supported by the accompanying evidence: A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior. Multiple tenants operate within their allocations while a full-device benchmark remains unaffected.

B:

Map VLAN or VRF boundaries, workload identities, firewall rules, east-west flows and attempted cross-tenant connections. It would not provide the evidence needed for the paired diagnosis and follow-up: A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. Check firmware, drivers, hypervisor or container runtime compatibility, partition inventory and isolation tests. Multiple tenants operate within their allocations while a full-device benchmark remains unaffected.

D:

Observe GPU idle gaps, CPU saturation, page faults, dataloader queue depth, NUMA affinity and PCIe locality. It would not provide the evidence needed for the paired diagnosis and follow-up: A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. Check firmware, drivers, hypervisor or container runtime compatibility, partition inventory and isolation tests. Multiple tenants operate within their allocations while a full-device benchmark remains unaffected.

Question 195

The acceptance board must decide whether an AI-infrastructure response to accelerator virtualization support is fit for controlled operation. The fictional operating facts are: A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. Signoff cannot rely on a good-looking utilization graph, a change ticket alone or an unmeasured promise. The response must include an appropriate governing engineering principle, a scoped intervention and a separate measurable acceptance criterion. Choose EXACTLY THREE statements that meet those roles.

  1. Multiple tenants operate within their allocations while a full-device benchmark remains unaffected.
  2. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior.
  3. GPU partitioning and virtualization require platform-licensed, driver-compatible capabilities verified for the target workload.
  4. The same-load measurement demonstrates reduced facility overhead without harming service performance.
  5. Apply least-privilege segmentation and test allowed model/data flows plus blocked management and tenant access.
  6. Storage capacity plans require dataset lineage, scratch amplification, checkpoint retention and growth forecasts, not initial dataset size only.

Explanation

Correct Answer

A:

A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. Multiple tenants operate within their allocations while a full-device benchmark remains unaffected. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior.

B:

A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior. GPU partitioning and virtualization require platform-licensed, driver-compatible capabilities verified for the target workload.

C:

A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. GPU partitioning and virtualization require platform-licensed, driver-compatible capabilities verified for the target workload. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior.

Incorrect Answers

D:

The same-load measurement demonstrates reduced facility overhead without harming service performance. It would not independently demonstrate that the control works: A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. Multiple tenants operate within their allocations while a full-device benchmark remains unaffected. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior.

E:

Apply least-privilege segmentation and test allowed model/data flows plus blocked management and tenant access. It does not implement the governing rule for this approval: A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior. Multiple tenants operate within their allocations while a full-device benchmark remains unaffected.

F:

Storage capacity plans require dataset lineage, scratch amplification, checkpoint retention and growth forecasts, not initial dataset size only. It is not the rule that supports the proposed change and acceptance test: A college seeks to divide a shared GPU pool among small inferencers without checking device partition support. GPU partitioning and virtualization require platform-licensed, driver-compatible capabilities verified for the target workload. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior.

Question 196

A data-center AI architecture committee considers accelerator virtualization support in its AI Infrastructure Components and Architecture review. At the affected site: A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. The committee is deliberately not selecting a vendor product or applying a live change yet. It needs the foundational reasoning that makes the next design decision defensible. Which ONE engineering principle directly governs this situation? Choose ONE answer.

  1. GPU partitioning and virtualization require platform-licensed, driver-compatible capabilities verified for the target workload.
  2. AI thermal designs must handle rack heat flux and liquid or air-cooling capacity under worst-case workloads.
  3. A cluster should tolerate a single link or fabric-device failure using validated redundant paths and routing convergence.
  4. AI servers, firmware, NIC drivers, GPU stack and fabric software form a versioned support matrix.

Explanation

Correct Answer

A:

A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. GPU partitioning and virtualization require platform-licensed, driver-compatible capabilities verified for the target workload. Check firmware, drivers, hypervisor or container runtime compatibility, partition inventory and isolation tests.

Incorrect Answers

B:

AI thermal designs must handle rack heat flux and liquid or air-cooling capacity under worst-case workloads. It does not govern the condition described: A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. GPU partitioning and virtualization require platform-licensed, driver-compatible capabilities verified for the target workload.

C:

A cluster should tolerate a single link or fabric-device failure using validated redundant paths and routing convergence. It does not govern the condition described: A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. GPU partitioning and virtualization require platform-licensed, driver-compatible capabilities verified for the target workload.

D:

AI servers, firmware, NIC drivers, GPU stack and fabric software form a versioned support matrix. It does not govern the condition described: A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. GPU partitioning and virtualization require platform-licensed, driver-compatible capabilities verified for the target workload.

Question 197

An infrastructure diagnostician is investigating accelerator virtualization support for the AI Infrastructure Components and Architecture team. The reported incident reads: A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. Competing teams attribute the symptoms to different layers of the AI stack. Before accepting a hypothesis or touching production settings, the investigator must choose a discriminating set of measurements. Which ONE evidence package is MOST relevant? Choose ONE answer.

  1. Inventory immutable datasets, intermediate artifacts, checkpoint rotations, compression and quota high-water marks.
  2. Check firmware, drivers, hypervisor or container runtime compatibility, partition inventory and isolation tests.
  3. Measure per-worker read concurrency, storage-node load, cache hit rates, queue depth and samples-per-second.
  4. Map VLAN or VRF boundaries, workload identities, firewall rules, east-west flows and attempted cross-tenant connections.

Explanation

Correct Answer

B:

A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. Check firmware, drivers, hypervisor or container runtime compatibility, partition inventory and isolation tests. GPU partitioning and virtualization require platform-licensed, driver-compatible capabilities verified for the target workload.

Incorrect Answers

A:

Inventory immutable datasets, intermediate artifacts, checkpoint rotations, compression and quota high-water marks. That check would not establish the decisive facts for this condition: A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. Check firmware, drivers, hypervisor or container runtime compatibility, partition inventory and isolation tests.

C:

Measure per-worker read concurrency, storage-node load, cache hit rates, queue depth and samples-per-second. That check would not establish the decisive facts for this condition: A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. Check firmware, drivers, hypervisor or container runtime compatibility, partition inventory and isolation tests.

D:

Map VLAN or VRF boundaries, workload identities, firewall rules, east-west flows and attempted cross-tenant connections. That check would not establish the decisive facts for this condition: A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. Check firmware, drivers, hypervisor or container runtime compatibility, partition inventory and isolation tests.

Question 198

During a bounded AI-platform change-control review concerning accelerator virtualization support, the site owner reports: A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. The proposed work must have a narrow implementation scope, first be validated against the failure evidence and avoid unrelated system-wide tuning. Which ONE intervention should the engineers pilot before any wider rollout? Choose ONE answer.

  1. Benchmark representative dataset shards and tune supported striping, caching or storage tiers based on bottlenecks.
  2. Size and balance the leaf-spine fabric for the observed communication pattern before adding accelerator nodes.
  3. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior.
  4. Validate and stage a supported end-to-end software bill of materials with rollback images.

Explanation

Correct Answer

C:

A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior. Check firmware, drivers, hypervisor or container runtime compatibility, partition inventory and isolation tests.

Incorrect Answers

A:

Benchmark representative dataset shards and tune supported striping, caching or storage tiers based on bottlenecks. That change would leave the reported problem unresolved: A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior.

B:

Size and balance the leaf-spine fabric for the observed communication pattern before adding accelerator nodes. That change would leave the reported problem unresolved: A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior.

D:

Validate and stage a supported end-to-end software bill of materials with rollback images. That change would leave the reported problem unresolved: A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior.

Question 199

Two independent technical reviewers examine a AI Infrastructure Components and Architecture problem involving accelerator virtualization support. Their shared incident log states: A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. One reviewer requires measurements capable of separating the observed problem from neighboring failure modes; the other requires a bounded action that follows those measurements. Choose EXACTLY TWO statements that together supply those two parts.

  1. Check firmware, drivers, hypervisor or container runtime compatibility, partition inventory and isolation tests.
  2. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior.
  3. Coordinate facility-validated cooling distribution and thermal controls before increasing sustained GPU power.
  4. Measure per-worker read concurrency, storage-node load, cache hit rates, queue depth and samples-per-second.
  5. Count metadata operations, open latency, namespace contention and worker-level file listing times.

Explanation

Correct Answer

A:

A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. Check firmware, drivers, hypervisor or container runtime compatibility, partition inventory and isolation tests. Multiple tenants operate within their allocations while a full-device benchmark remains unaffected.

B:

A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior. Multiple tenants operate within their allocations while a full-device benchmark remains unaffected.

Incorrect Answers

C:

Coordinate facility-validated cooling distribution and thermal controls before increasing sustained GPU power. It is not the scoped response supported by the accompanying evidence: A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior. Multiple tenants operate within their allocations while a full-device benchmark remains unaffected.

D:

Measure per-worker read concurrency, storage-node load, cache hit rates, queue depth and samples-per-second. It would not provide the evidence needed for the paired diagnosis and follow-up: A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. Check firmware, drivers, hypervisor or container runtime compatibility, partition inventory and isolation tests. Multiple tenants operate within their allocations while a full-device benchmark remains unaffected.

E:

Count metadata operations, open latency, namespace contention and worker-level file listing times. It would not provide the evidence needed for the paired diagnosis and follow-up: A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. Check firmware, drivers, hypervisor or container runtime compatibility, partition inventory and isolation tests. Multiple tenants operate within their allocations while a full-device benchmark remains unaffected.

Question 200

The acceptance board must decide whether an AI-infrastructure response to accelerator virtualization support is fit for controlled operation. The fictional operating facts are: A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. Signoff cannot rely on a good-looking utilization graph, a change ticket alone or an unmeasured promise. The response must include an appropriate governing engineering principle, a scoped intervention and a separate measurable acceptance criterion. Choose EXACTLY THREE statements that meet those roles.

  1. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior.
  2. High-density GPU racks must fit measured branch-circuit, PDU and failover electrical budgets under realistic power transients.
  3. Reserve appropriate service classes and relieve burst contention after validating which path contributes to tail delay.
  4. Multiple tenants operate within their allocations while a full-device benchmark remains unaffected.
  5. GPU partitioning and virtualization require platform-licensed, driver-compatible capabilities verified for the target workload.
  6. Authorized training jobs complete while isolation probes cannot reach disallowed resources.

Explanation

Correct Answer

A:

A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior. GPU partitioning and virtualization require platform-licensed, driver-compatible capabilities verified for the target workload.

D:

A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. Multiple tenants operate within their allocations while a full-device benchmark remains unaffected. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior.

E:

A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. GPU partitioning and virtualization require platform-licensed, driver-compatible capabilities verified for the target workload. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior.

Incorrect Answers

B:

High-density GPU racks must fit measured branch-circuit, PDU and failover electrical budgets under realistic power transients. It is not the rule that supports the proposed change and acceptance test: A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. GPU partitioning and virtualization require platform-licensed, driver-compatible capabilities verified for the target workload. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior.

C:

Reserve appropriate service classes and relieve burst contention after validating which path contributes to tail delay. It does not implement the governing rule for this approval: A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior. Multiple tenants operate within their allocations while a full-device benchmark remains unaffected.

F:

Authorized training jobs complete while isolation probes cannot reach disallowed resources. It would not independently demonstrate that the control works: A retail cluster enables an experimental virtual GPU mode and distributed workers fail to see consistent device topology. Multiple tenants operate within their allocations while a full-device benchmark remains unaffected. Adopt only vendor-supported accelerator-sharing modes and reserve dedicated resources for jobs that need full-device behavior.

Implementing Cisco Data Center AI Infrastructure (DCAI) (300-640) Practice Tests