A reports an inadequate or failed acceptance condition: Capacity Reservation charges remain close to On-Demand despite the desired Spot savings. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that the fleet drains interrupted work and restarts it without losing job progress. The underlying design requirement is spot instances discount interruptible capacity; unrecoverable stateful primary databases are poor candidates.

C reports an inadequate or failed acceptance condition: All jobs run On-Demand at the old rate even though their interruption tolerance is unchanged. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that the fleet drains interrupted work and restarts it without losing job progress. The underlying design requirement is spot instances discount interruptible capacity; unrecoverable stateful primary databases are poor candidates.

D reports an inadequate or failed acceptance condition: One Spot pool interruption loses every in-flight job because no checkpoint or retry exists. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that the fleet drains interrupted work and restarts it without losing job progress. The underlying design requirement is spot instances discount interruptible capacity; unrecoverable stateful primary databases are poor candidates.

A would place S3 gateway endpoints in unrelated VPC route tables and expect this VPC's NAT charges to fall. This is not a defensible answer to the stated constraint (Costs spike because each private subnet uses its own NAT gateway for S3 bucket reads): s3 gateway endpoint data paths avoid NAT processing charges for eligible S3 access. The required design is introduce S3 gateway endpoints in the matching VPC route tables and retain NAT for needed public egress; acceptance depends on evidence that nAT processed bytes drop for S3 traffic while private S3 requests still work.

B would route public internet egress through an S3 gateway endpoint after removing the required NAT path. This is not a defensible answer to the stated constraint (Costs spike because each private subnet uses its own NAT gateway for S3 bucket reads): s3 gateway endpoint data paths avoid NAT processing charges for eligible S3 access. The required design is introduce S3 gateway endpoints in the matching VPC route tables and retain NAT for needed public egress; acceptance depends on evidence that nAT processed bytes drop for S3 traffic while private S3 requests still work.

D would increase NAT gateway count solely to carry private S3 bucket traffic. This is not a defensible answer to the stated constraint (Costs spike because each private subnet uses its own NAT gateway for S3 bucket reads): s3 gateway endpoint data paths avoid NAT processing charges for eligible S3 access. The required design is introduce S3 gateway endpoints in the matching VPC route tables and retain NAT for needed public egress; acceptance depends on evidence that nAT processed bytes drop for S3 traffic while private S3 requests still work.

A reports an inadequate or failed acceptance condition: An S3 gateway endpoint is available in another VPC but this VPC still sends S3 traffic via NAT. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that nAT processed bytes drop for S3 traffic while private S3 requests still work. The underlying design requirement is s3 gateway endpoint data paths avoid NAT processing charges for eligible S3 access.

B reports an inadequate or failed acceptance condition: Public internet egress fails after NAT is removed, even though S3 endpoint traffic works. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that nAT processed bytes drop for S3 traffic while private S3 requests still work. The underlying design requirement is s3 gateway endpoint data paths avoid NAT processing charges for eligible S3 access.

C reports an inadequate or failed acceptance condition: Additional NAT gateways carry increased S3 bytes without a reduction in processed network traffic. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that nAT processed bytes drop for S3 traffic while private S3 requests still work. The underlying design requirement is s3 gateway endpoint data paths avoid NAT processing charges for eligible S3 access.

B would review Cost Explorer manually at month-end rather than receiving advance budget alerts. This is not a defensible answer to the stated constraint (A company wants advance notification when the monthly spending trend exceeds an approved budget): budgets monitors spend and forecast thresholds; Cost Explorer alone does not enforce budget alerts. The required design is create AWS Budgets thresholds and forecasts with configured notifications; acceptance depends on evidence that budget alerts arrive at projected or actual spending thresholds.

C would create CloudWatch instance CPU alarms to forecast all account spending. This is not a defensible answer to the stated constraint (A company wants advance notification when the monthly spending trend exceeds an approved budget): budgets monitors spend and forecast thresholds; Cost Explorer alone does not enforce budget alerts. The required design is create AWS Budgets thresholds and forecasts with configured notifications; acceptance depends on evidence that budget alerts arrive at projected or actual spending thresholds.

D would configure AWS Budgets with no thresholds, recipients, or forecast notification rules. This is not a defensible answer to the stated constraint (A company wants advance notification when the monthly spending trend exceeds an approved budget): budgets monitors spend and forecast thresholds; Cost Explorer alone does not enforce budget alerts. The required design is create AWS Budgets thresholds and forecasts with configured notifications; acceptance depends on evidence that budget alerts arrive at projected or actual spending thresholds.

A reports an inadequate or failed acceptance condition: The team discovers the overrun during month-end manual billing review. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that budget alerts arrive at projected or actual spending thresholds. The underlying design requirement is budgets monitors spend and forecast thresholds; Cost Explorer alone does not enforce budget alerts.

C reports an inadequate or failed acceptance condition: A CPU CloudWatch alarm fires but monthly account spending already exceeds the budget. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that budget alerts arrive at projected or actual spending thresholds. The underlying design requirement is budgets monitors spend and forecast thresholds; Cost Explorer alone does not enforce budget alerts.

D reports an inadequate or failed acceptance condition: An AWS Budget exists but no forecast threshold triggers a notification. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that budget alerts arrive at projected or actual spending thresholds. The underlying design requirement is budgets monitors spend and forecast thresholds; Cost Explorer alone does not enforce budget alerts.

A would move every S3 object immediately into Glacier Deep Archive despite unpredictable retrievals. This is not a defensible answer to the stated constraint (A data lake stores years of infrequently accessed S3 objects with unpredictable retrieval intervals): intelligent-Tiering moves eligible objects between access tiers automatically, unlike manual guessing of object ages. The required design is choose a suitable S3 Intelligent-Tiering configuration where object access patterns justify monitoring charges; acceptance depends on evidence that storage class transitions reflect actual usage and total cost improves after monitoring fees.

B would leave all objects in S3 Standard indefinitely without testing access-frequency or cost patterns. This is not a defensible answer to the stated constraint (A data lake stores years of infrequently accessed S3 objects with unpredictable retrieval intervals): intelligent-Tiering moves eligible objects between access tiers automatically, unlike manual guessing of object ages. The required design is choose a suitable S3 Intelligent-Tiering configuration where object access patterns justify monitoring charges; acceptance depends on evidence that storage class transitions reflect actual usage and total cost improves after monitoring fees.

D would select Intelligent-Tiering without including monitoring charges in the cost comparison. This is not a defensible answer to the stated constraint (A data lake stores years of infrequently accessed S3 objects with unpredictable retrieval intervals): intelligent-Tiering moves eligible objects between access tiers automatically, unlike manual guessing of object ages. The required design is choose a suitable S3 Intelligent-Tiering configuration where object access patterns justify monitoring charges; acceptance depends on evidence that storage class transitions reflect actual usage and total cost improves after monitoring fees.

A reports an inadequate or failed acceptance condition: Deep Archive retrieval of a needed object takes much longer than its unpredictable access requirement allows. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that storage class transitions reflect actual usage and total cost improves after monitoring fees. The underlying design requirement is intelligent-Tiering moves eligible objects between access tiers automatically, unlike manual guessing of object ages.

B reports an inadequate or failed acceptance condition: All old objects remain S3 Standard despite long-term rare access. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that storage class transitions reflect actual usage and total cost improves after monitoring fees. The underlying design requirement is intelligent-Tiering moves eligible objects between access tiers automatically, unlike manual guessing of object ages.

C reports an inadequate or failed acceptance condition: Monitoring fees plus transitions make Intelligent-Tiering more expensive than the measured alternative. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that storage class transitions reflect actual usage and total cost improves after monitoring fees. The underlying design requirement is intelligent-Tiering moves eligible objects between access tiers automatically, unlike manual guessing of object ages.

B would query CloudWatch CPU metrics and assume they yield resource-level cost allocation and commitments. This is not a defensible answer to the stated constraint (An analytics department needs detailed per-resource daily cost investigation including amortized commitments): cUR offers richer granular billing datasets than summary dashboard totals. The required design is enable Cost and Usage Reports with the necessary resource IDs and query the files with Athena; acceptance depends on evidence that athena queries return itemized usage and amortized charge columns by team.

C would generate CUR without enabling resource identifiers or considering how amortized discounts appear. This is not a defensible answer to the stated constraint (An analytics department needs detailed per-resource daily cost investigation including amortized commitments): cUR offers richer granular billing datasets than summary dashboard totals. The required design is enable Cost and Usage Reports with the necessary resource IDs and query the files with Athena; acceptance depends on evidence that athena queries return itemized usage and amortized charge columns by team.

D would use only the monthly billing summary, which lacks detailed amortized resource line items. This is not a defensible answer to the stated constraint (An analytics department needs detailed per-resource daily cost investigation including amortized commitments): cUR offers richer granular billing datasets than summary dashboard totals. The required design is enable Cost and Usage Reports with the necessary resource IDs and query the files with Athena; acceptance depends on evidence that athena queries return itemized usage and amortized charge columns by team.

A reports an inadequate or failed acceptance condition: An EC2 CPU metric shows a spike but contains no billed usage line or commitment amortization. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that athena queries return itemized usage and amortized charge columns by team. The underlying design requirement is cUR offers richer granular billing datasets than summary dashboard totals.

C reports an inadequate or failed acceptance condition: CUR lacks resource IDs, so SQL cannot attribute costs to each requested workload. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that athena queries return itemized usage and amortized charge columns by team. The underlying design requirement is cUR offers richer granular billing datasets than summary dashboard totals.

D reports an inadequate or failed acceptance condition: A bill summary shows monthly totals but cannot attribute amortized charges to a resource. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that athena queries return itemized usage and amortized charge columns by team. The underlying design requirement is cUR offers richer granular billing datasets than summary dashboard totals.

A would buy larger instances for the 6% utilization workload without evaluating optimizer recommendations. This is not a defensible answer to the stated constraint (Expensive Graviton-ineligible compute instances run at 6% CPU for months): right-sizing must preserve memory, architecture, licenses and workload latency constraints. The required design is review Compute Optimizer evidence and rightsize to an application-compatible smaller instance family; acceptance depends on evidence that cloudWatch latency and capacity remain within SLO after utilization and cost decline.

B would move the application to Graviton instances before checking architecture compatibility. This is not a defensible answer to the stated constraint (Expensive Graviton-ineligible compute instances run at 6% CPU for months): right-sizing must preserve memory, architecture, licenses and workload latency constraints. The required design is review Compute Optimizer evidence and rightsize to an application-compatible smaller instance family; acceptance depends on evidence that cloudWatch latency and capacity remain within SLO after utilization and cost decline.

D would right-size solely on average CPU while ignoring memory, burst, network, and SLO limits. This is not a defensible answer to the stated constraint (Expensive Graviton-ineligible compute instances run at 6% CPU for months): right-sizing must preserve memory, architecture, licenses and workload latency constraints. The required design is review Compute Optimizer evidence and rightsize to an application-compatible smaller instance family; acceptance depends on evidence that cloudWatch latency and capacity remain within SLO after utilization and cost decline.

A reports an inadequate or failed acceptance condition: The new instance is bigger and its hourly cost rises although original CPU was only 6%. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that cloudWatch latency and capacity remain within SLO after utilization and cost decline. The underlying design requirement is right-sizing must preserve memory, architecture, licenses and workload latency constraints.

B reports an inadequate or failed acceptance condition: A Graviton instance boot test fails because the compiled application is architecture-incompatible. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that cloudWatch latency and capacity remain within SLO after utilization and cost decline. The underlying design requirement is right-sizing must preserve memory, architecture, licenses and workload latency constraints.

C reports an inadequate or failed acceptance condition: Average CPU declines but memory saturation or request latency exceeds service SLO. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that cloudWatch latency and capacity remain within SLO after utilization and cost decline. The underlying design requirement is right-sizing must preserve memory, architecture, licenses and workload latency constraints.