100% Latest & Updated Google Professional Cloud Architect Practice Test Questions, Exam Dumps & Verified Answers!
30 Days Free Updates, Instant Download!
Question 60
A news publisher stores immutable media assets that must survive data-center failures and are accessed as objects. Which operational action best maintains cost-aware media retention after launch?
- Move all objects to the coldest class irrespective of usage
- Apply lifecycle classes only after checking retrieval-frequency and recovery needs
- Keep indefinite unused versions without review
- Review only VMs and neglect object storage request/billing patterns
Explanation
Correct Answer
B:Storage-tier savings must account for access costs and retrieval behavior
Incorrect Answers
A:Can increase retrieval charges and latency. Storage-tier savings must account for access costs and retrieval behavior.
C:Increases storage cost. Storage-tier savings must account for access costs and retrieval behavior.
D:Cannot optimize media retention. Storage-tier savings must account for access costs and retrieval behavior.
Question 61
A scientific modeling team needs sustained accelerator access for distributed training with custom container runtimes. Which architecture most directly meets the need for orchestrated accelerator workloads?
- Use GKE with appropriately sized accelerator-enabled nodes or a suitable managed training platform
- Schedule all distributed accelerator training on CPU-only Cloud Run services
- Run every training worker on a fixed general-purpose VM without orchestration
- Use BigQuery for GPU container scheduling
Explanation
Correct Answer
A:Workload requirements determine whether granular Kubernetes control is justified
Incorrect Answers
B:Does not expose the requested accelerator capacity. Workload requirements determine whether granular Kubernetes control is justified.
C:Complicates distributed scheduling and recovery. Workload requirements determine whether granular Kubernetes control is justified.
D:A SQL warehouse is not a cluster scheduler. Workload requirements determine whether granular Kubernetes control is justified.
Question 62
A scientific modeling team needs sustained accelerator access for distributed training with custom container runtimes. Which implementation choice most directly delivers accelerator scheduling isolation?
- Allow accelerator pods without resource requests
- Configure appropriate GPU/TPU resource requests and dedicated capacity scheduling
- Treat a TPU request as identical to every GPU runtime requirement
- Place all distributed workers in a container image but no shared job topology
Explanation
Correct Answer
B:Explicit accelerator scheduling avoids oversubscription and accidental CPU-only placement
Incorrect Answers
A:Scheduler cannot reliably reserve requested devices. Explicit accelerator scheduling avoids oversubscription and accidental CPU-only placement.
C:Framework and hardware compatibility differ. Explicit accelerator scheduling avoids oversubscription and accidental CPU-only placement.
D:Does not coordinate accelerator workloads. Explicit accelerator scheduling avoids oversubscription and accidental CPU-only placement.
Question 63
A scientific modeling team needs sustained accelerator access for distributed training with custom container runtimes. Which validation would best establish evidence that training scales efficiently?
- Check only node count after scheduling
- Measure accelerator utilization and step throughput during a representative distributed run
- Measure storage disk size as the performance benchmark
- Compare one CPU training step to idle accelerator capacity
Explanation
Correct Answer
B:Provisioned accelerators alone are not evidence of useful model training throughput
Incorrect Answers
A:Does not show devices are doing training work. Provisioned accelerators alone are not evidence of useful model training throughput.
C:Not model throughput. Provisioned accelerators alone are not evidence of useful model training throughput.
D:Does not represent scaled workload performance. Provisioned accelerators alone are not evidence of useful model training throughput.
Question 64
A scientific modeling team needs sustained accelerator access for distributed training with custom container runtimes. Which operational action best maintains ongoing accelerator efficiency after launch?
- Add accelerators whenever total training spend rises
- Monitor only free CPU cores
- Right-size node pools based on utilization and queue wait times
- Ignore queue wait times while right-sizing
Explanation
Correct Answer
C:Idle expensive accelerators can dominate model-training cost
Incorrect Answers
A:Can worsen cost without throughput gains. Idle expensive accelerators can dominate model-training cost.
B:Misses accelerator idleness. Idle expensive accelerators can dominate model-training cost.
D:Can make tasks unable to obtain required devices. Idle expensive accelerators can dominate model-training cost.
Question 65
A manufacturer runs Kubernetes on premises and must connect private factory APIs to Google Cloud services without exposing them publicly. Which architecture most directly meets the need for hybrid private network and cluster integration?
- Expose every factory API through public VM IPs instead of private routes
- Join factory and cloud networks with overlapping address ranges
- Use only application DNS names without establishing connectivity
- Use private connectivity and a documented VPC/GKE network topology for permitted factory flows
Explanation
Correct Answer
D:Explicit routes and private addresses preserve intended trust boundaries across environments
Incorrect Answers
A:Violates the intended private access boundary. Explicit routes and private addresses preserve intended trust boundaries across environments.
B:Routing would be ambiguous. Explicit routes and private addresses preserve intended trust boundaries across environments.
C:Name resolution cannot substitute for reachable private paths. Explicit routes and private addresses preserve intended trust boundaries across environments.
Question 66
A manufacturer runs Kubernetes on premises and must connect private factory APIs to Google Cloud services without exposing them publicly. Which implementation choice most directly delivers restricted hybrid service access?
- Permit all on-premises subnets to reach all cloud services
- Rely on RFC1918 addressing without authenticating workloads
- Use narrowly scoped firewall and identity controls on authorized private service paths
- Disable cloud firewall logging on the hybrid edge
Explanation
Correct Answer
C:A private route alone is not proof that a factory workload is authorized
Incorrect Answers
A:Overbroad network rules violate least privilege. A private route alone is not proof that a factory workload is authorized.
B:Private IP alone is not identity. A private route alone is not proof that a factory workload is authorized.
D:Hinders boundary troubleshooting. A private route alone is not proof that a factory workload is authorized.
Question 67
A manufacturer runs Kubernetes on premises and must connect private factory APIs to Google Cloud services without exposing them publicly. Which validation would best establish validation of hybrid network boundaries?
- Test approved factory-to-cloud requests and denied unauthorized subnet/service combinations
- Ping one VM from the factory network
- Test only a valid client and omit negative flows
- Examine only IPAM inventory
Explanation
Correct Answer
A:Positive and negative paths detect overly broad firewall and identity policy
Incorrect Answers
B:Does not validate service identity or denied paths. Positive and negative paths detect overly broad firewall and identity policy.
C:Overly broad access can go undetected. Positive and negative paths detect overly broad firewall and identity policy.
D:Does not establish live routing or access behavior. Positive and negative paths detect overly broad firewall and identity policy.
Question 68
A manufacturer runs Kubernetes on premises and must connect private factory APIs to Google Cloud services without exposing them publicly. Which operational action best maintains change-safe hybrid operations after launch?
- Freeze hybrid routing policy despite factory subnet changes
- Monitor private-link route health and version controlled network policy changes
- Monitor only cloud egress billing
- Permit emergency firewall rules indefinitely without review
Explanation
Correct Answer
B:Hybrid connectivity can drift when either side changes routes or service identities
Incorrect Answers
A:Leaves stale routes. Hybrid connectivity can drift when either side changes routes or service identities.
C:Does not show private link path health. Hybrid connectivity can drift when either side changes routes or service identities.
D:Creates persistent broad trust paths. Hybrid connectivity can drift when either side changes routes or service identities.
Question 69
A bank inventories 300 dependent applications before a time-constrained data-center exit; migration order is disputed. Which architecture most directly meets the need for fact-based migration wave planning?
- Migrate 300 apps in alphabetical server order
- Migrate the largest database first without checking its consumers
- Choose migration waves by VM CPU utilization only
- Use Migration Center assessment and dependency mapping to define application migration waves
Explanation
Correct Answer
D:Dependency evidence reduces outages caused by moving a supporting service after its consumer
Incorrect Answers
A:Ignores dependencies and business criticality. Dependency evidence reduces outages caused by moving a supporting service after its consumer.
B:Can break applications still on-premises. Dependency evidence reduces outages caused by moving a supporting service after its consumer.
C:Cannot represent application coupling. Dependency evidence reduces outages caused by moving a supporting service after its consumer.
Question 70
A bank inventories 300 dependent applications before a time-constrained data-center exit; migration order is disputed. Which implementation choice most directly delivers controlled pilot migration?
- Move a critical interdependent payment suite as the first untested wave
- Skip rollback planning because rehosted VMs are unchanged
- Move a small dependent application group with documented cutover and rollback steps
- Disable all app telemetry during migration to reduce costs
Explanation
Correct Answer
C:Pilots expose unknown network, authentication, and licensing dependencies early
Incorrect Answers
A:Maximizes blast radius. Pilots expose unknown network, authentication, and licensing dependencies early.
B:Network and identity integration still change. Pilots expose unknown network, authentication, and licensing dependencies early.
D:Removes evidence when dependencies fail. Pilots expose unknown network, authentication, and licensing dependencies early.
Google Cloud Certified - Professional Cloud Architect (Professional Cloud Architect) Practice Tests