A reports an inadequate or failed acceptance condition: Distributed workers all block on a shared lock, so completion time does not improve. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that worker count increases with backlog and completion time falls without shared-lock contention. The underlying design requirement is queue-driven parallelism works when work units are independent and consumers safely handle retries.

B reports an inadequate or failed acceptance condition: Multiple workers acknowledge the same task or lose tasks because queue ownership is not durable. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that worker count increases with backlog and completion time falls without shared-lock contention. The underlying design requirement is queue-driven parallelism works when work units are independent and consumers safely handle retries.

C reports an inadequate or failed acceptance condition: One large EC2 instance remains the bottleneck for every independent batch work unit. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that worker count increases with backlog and completion time falls without shared-lock contention. The underlying design requirement is queue-driven parallelism works when work units are independent and consumers safely handle retries.

B would move the database to S3 Standard and expect block-level random write IOPS like EBS. This is not a defensible answer to the stated constraint (A latency-sensitive SQL Server database has high random IOPS but EBS metrics show an insufficient provisioned level): both volume performance and EC2 attachment bandwidth cap EBS throughput. The required design is benchmark and select a supported EBS provisioned-IOPS volume/instance throughput combination; acceptance depends on evidence that p99 storage latency improves and provisioned IOPS are actually attainable on the instance.

C would increase SQL Server instance CPU count without evaluating the EBS IOPS limit. This is not a defensible answer to the stated constraint (A latency-sensitive SQL Server database has high random IOPS but EBS metrics show an insufficient provisioned level): both volume performance and EC2 attachment bandwidth cap EBS throughput. The required design is benchmark and select a supported EBS provisioned-IOPS volume/instance throughput combination; acceptance depends on evidence that p99 storage latency improves and provisioned IOPS are actually attainable on the instance.

D would provision higher EBS IOPS than the instance can deliver and assume measured throughput rises. This is not a defensible answer to the stated constraint (A latency-sensitive SQL Server database has high random IOPS but EBS metrics show an insufficient provisioned level): both volume performance and EC2 attachment bandwidth cap EBS throughput. The required design is benchmark and select a supported EBS provisioned-IOPS volume/instance throughput combination; acceptance depends on evidence that p99 storage latency improves and provisioned IOPS are actually attainable on the instance.

A reports an inadequate or failed acceptance condition: S3 GET requests succeed but the SQL database still lacks random-write EBS throughput. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that p99 storage latency improves and provisioned IOPS are actually attainable on the instance. The underlying design requirement is both volume performance and EC2 attachment bandwidth cap EBS throughput.

C reports an inadequate or failed acceptance condition: EC2 CPU decreases but SQL storage I/O latency remains above the threshold. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that p99 storage latency improves and provisioned IOPS are actually attainable on the instance. The underlying design requirement is both volume performance and EC2 attachment bandwidth cap EBS throughput.

D reports an inadequate or failed acceptance condition: The EBS volume advertises provisioned IOPS that exceed the instance attachment cap. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that p99 storage latency improves and provisioned IOPS are actually attainable on the instance. The underlying design requirement is both volume performance and EC2 attachment bandwidth cap EBS throughput.

A would keep synchronous transactional calls crossing Regions and increase EC2 CPU alone. This is not a defensible answer to the stated constraint (A transactional app makes dozens of cross-Region synchronous database calls per user request): cross-Region network RTT accumulates in synchronous call chains even when individual services are fast. The required design is move latency-sensitive data and compute into one appropriate Region or redesign asynchronous integration; acceptance depends on evidence that end-to-end traces show fewer interregional round trips and improved user p95 latency.

B would put only CloudFront in front of the application while leaving back-end synchronous cross-Region calls unchanged. This is not a defensible answer to the stated constraint (A transactional app makes dozens of cross-Region synchronous database calls per user request): cross-Region network RTT accumulates in synchronous call chains even when individual services are fast. The required design is move latency-sensitive data and compute into one appropriate Region or redesign asynchronous integration; acceptance depends on evidence that end-to-end traces show fewer interregional round trips and improved user p95 latency.

D would increase database connection timeouts to hide rather than reduce cross-Region round trips. This is not a defensible answer to the stated constraint (A transactional app makes dozens of cross-Region synchronous database calls per user request): cross-Region network RTT accumulates in synchronous call chains even when individual services are fast. The required design is move latency-sensitive data and compute into one appropriate Region or redesign asynchronous integration; acceptance depends on evidence that end-to-end traces show fewer interregional round trips and improved user p95 latency.

A reports an inadequate or failed acceptance condition: Server CPU is low yet traces still contain dozens of synchronous inter-Region SQL calls. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that end-to-end traces show fewer interregional round trips and improved user p95 latency. The underlying design requirement is cross-Region network RTT accumulates in synchronous call chains even when individual services are fast.

B reports an inadequate or failed acceptance condition: CloudFront edge hits increase but the user request still blocks on remote database calls. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that end-to-end traces show fewer interregional round trips and improved user p95 latency. The underlying design requirement is cross-Region network RTT accumulates in synchronous call chains even when individual services are fast.

C reports an inadequate or failed acceptance condition: A higher timeout reduces errors but the p95 response time remains above the objective. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that end-to-end traces show fewer interregional round trips and improved user p95 latency. The underlying design requirement is cross-Region network RTT accumulates in synchronous call chains even when individual services are fast.

B would use On-Demand Capacity Reservations as if they inherently give Savings Plan purchase discounts. This is not a defensible answer to the stated constraint (A steady application fleet uses 24/7 EC2 instances but periodically changes instance families): compute Savings Plans discount flexible compute; Reserved Instances can be less flexible for changing families. The required design is use Compute Savings Plans for the stable eligible compute baseline after coverage analysis; acceptance depends on evidence that commitment coverage and utilization improve without constraining every workload to one family.

C would commit to hourly spend above the persistent eligible baseline without evaluating utilization risk. This is not a defensible answer to the stated constraint (A steady application fleet uses 24/7 EC2 instances but periodically changes instance families): compute Savings Plans discount flexible compute; Reserved Instances can be less flexible for changing families. The required design is use Compute Savings Plans for the stable eligible compute baseline after coverage analysis; acceptance depends on evidence that commitment coverage and utilization improve without constraining every workload to one family.

D would purchase one instance-family-specific Regional Reserved Instance commitment for every fluctuating family. This is not a defensible answer to the stated constraint (A steady application fleet uses 24/7 EC2 instances but periodically changes instance families): compute Savings Plans discount flexible compute; Reserved Instances can be less flexible for changing families. The required design is use Compute Savings Plans for the stable eligible compute baseline after coverage analysis; acceptance depends on evidence that commitment coverage and utilization improve without constraining every workload to one family.

A reports an inadequate or failed acceptance condition: A Capacity Reservation is Active but the compute bill retains ordinary On-Demand pricing. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that commitment coverage and utilization improve without constraining every workload to one family. The underlying design requirement is compute Savings Plans discount flexible compute; Reserved Instances can be less flexible for changing families.

C reports an inadequate or failed acceptance condition: Savings Plan unused commitment grows because actual eligible compute is below the purchased baseline. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that commitment coverage and utilization improve without constraining every workload to one family. The underlying design requirement is compute Savings Plans discount flexible compute; Reserved Instances can be less flexible for changing families.

D reports an inadequate or failed acceptance condition: Regional Reserved Instance coverage stays limited to one family despite the fleet's changing families. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that commitment coverage and utilization improve without constraining every workload to one family. The underlying design requirement is compute Savings Plans discount flexible compute; Reserved Instances can be less flexible for changing families.

A would store all intermediate scientific results only on ephemeral Spot disks with no retry persistence. This is not a defensible answer to the stated constraint (A spot-tolerant scientific workload spends most of its budget on short-lived stateless worker pools): spot reduces interruptible compute costs; checkpointing and diversification are required for resilience. The required design is use diversified EC2 Spot or mixed-instance policies with interruption-aware checkpointing; acceptance depends on evidence that jobs complete after spot interruptions while average instance spend declines.

B would use only long-term On-Demand capacity despite the stated interruption-tolerant worker pool. This is not a defensible answer to the stated constraint (A spot-tolerant scientific workload spends most of its budget on short-lived stateless worker pools): spot reduces interruptible compute costs; checkpointing and diversification are required for resilience. The required design is use diversified EC2 Spot or mixed-instance policies with interruption-aware checkpointing; acceptance depends on evidence that jobs complete after spot interruptions while average instance spend declines.

D would launch all batch jobs as Spot on one instance type and AZ with no checkpoint recovery. This is not a defensible answer to the stated constraint (A spot-tolerant scientific workload spends most of its budget on short-lived stateless worker pools): spot reduces interruptible compute costs; checkpointing and diversification are required for resilience. The required design is use diversified EC2 Spot or mixed-instance policies with interruption-aware checkpointing; acceptance depends on evidence that jobs complete after spot interruptions while average instance spend declines.

A reports an inadequate or failed acceptance condition: Spot nodes terminate and all unpersisted results from the batch are lost. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that jobs complete after spot interruptions while average instance spend declines. The underlying design requirement is spot reduces interruptible compute costs; checkpointing and diversification are required for resilience.

B reports an inadequate or failed acceptance condition: Workers run reliably On-Demand but the average instance spend stays unchanged. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that jobs complete after spot interruptions while average instance spend declines. The underlying design requirement is spot reduces interruptible compute costs; checkpointing and diversification are required for resilience.

C reports an inadequate or failed acceptance condition: A single Spot interruption aborts all scientific jobs with no checkpoint recovery. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that jobs complete after spot interruptions while average instance spend declines. The underlying design requirement is spot reduces interruptible compute costs; checkpointing and diversification are required for resilience.

B would keep infrequently read storage in S3 Standard forever without evaluating lifecycle transitions. This is not a defensible answer to the stated constraint (A storage lake contains old data with verified rare reads and long retention windows): lifecycle transitions reduce storage charges when access frequency and retrieval penalties justify the tier. The required design is apply suitable S3 Lifecycle transitions and object age policies after retrieval-cost analysis; acceptance depends on evidence that storage-class inventory matches aging policy and end-to-end retrieval cost stays within target.

C would move data to a minimum-duration archival class without modeling early-deletion and retrieval charges. This is not a defensible answer to the stated constraint (A storage lake contains old data with verified rare reads and long retention windows): lifecycle transitions reduce storage charges when access frequency and retrieval penalties justify the tier. The required design is apply suitable S3 Lifecycle transitions and object age policies after retrieval-cost analysis; acceptance depends on evidence that storage-class inventory matches aging policy and end-to-end retrieval cost stays within target.

D would configure immediate expiration of objects despite required long retention windows. This is not a defensible answer to the stated constraint (A storage lake contains old data with verified rare reads and long retention windows): lifecycle transitions reduce storage charges when access frequency and retrieval penalties justify the tier. The required design is apply suitable S3 Lifecycle transitions and object age policies after retrieval-cost analysis; acceptance depends on evidence that storage-class inventory matches aging policy and end-to-end retrieval cost stays within target.

A reports an inadequate or failed acceptance condition: Storage-class inventory shows long-lived objects never leave S3 Standard. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that storage-class inventory matches aging policy and end-to-end retrieval cost stays within target. The underlying design requirement is lifecycle transitions reduce storage charges when access frequency and retrieval penalties justify the tier.

C reports an inadequate or failed acceptance condition: Deep Archive early-deletion and retrieval charges exceed the projected class savings. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that storage-class inventory matches aging policy and end-to-end retrieval cost stays within target. The underlying design requirement is lifecycle transitions reduce storage charges when access frequency and retrieval penalties justify the tier.

D reports an inadequate or failed acceptance condition: Lifecycle configuration expires an object before the required retention date. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that storage-class inventory matches aging policy and end-to-end retrieval cost stays within target. The underlying design requirement is lifecycle transitions reduce storage charges when access frequency and retrieval penalties justify the tier.