Amazon AWS Certified Solutions Architect Associate SAA-C03 Compute Purchasing Utilization and Cost Optimization Practice Test

 

Topic 17 covers compute purchasing utilization and cost optimization for the AWS Certified Solutions Architect – Associate certification. These original practice questions apply the verified SAA-C03 objectives to practical decisions and troubleshooting. Select one answer unless a fixed number is requested. For broader preparation, visit the AWS Certified Solutions Architect Associate SAA-C03 Exam Dumps page. Each option includes an explanation of the relevant behavior and scenario constraints.

Question 1

A stateless production fleet has a stable 24×7 baseline across several EC2 instance families, with some use shifting to Fargate. Which discount model offers broad compute flexibility?

1. Use On-Demand while the short-lived or uncertain workload is characterized.
2. Use Spot Instances for the interruption-tolerant portion.
3. Use a Compute Savings Plan for the measured baseline commitment.
4. Evaluate an EC2 Instance Savings Plan or suitable Reserved Instance for the predictable constrained baseline.
5. Commit only the durable baseline and serve variable peaks with uncommitted or interruption-tolerant capacity as appropriate.

Correct Answer: C

 

Correct Answer

Answer C is correct because Compute Savings Plans apply flexibly across eligible EC2, Fargate, and Lambda usage rather than binding the commitment to one instance family. This directly meets the decisive requirement: Compute Savings Plans.

Incorrect Answers

Answer A is incorrect because On-Demand avoids an up-front term commitment and is appropriate when future steady usage is not yet established. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Compute Savings Plans.

Answer B is incorrect because Spot uses spare EC2 capacity at discounted rates but can be interrupted, so it fits checkpointed or fault-tolerant work. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Compute Savings Plans.

Answer D is incorrect because A stable, known EC2 usage pattern can trade flexibility for a larger discount where its family/Region constraints are acceptable. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Compute Savings Plans.

Answer E is incorrect because Buying a commitment sized to peak demand risks paying for capacity when the burst disappears. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Compute Savings Plans.

 

Question 2

A batch-rendering fleet can checkpoint work and tolerate interruptions. Which EC2 purchasing model can reduce cost most aggressively for the interruptible portion?

1. Commit only the durable baseline and serve variable peaks with uncommitted or interruption-tolerant capacity as appropriate.
2. Use Spot Instances for the interruption-tolerant portion.
3. Evaluate an EC2 Instance Savings Plan or suitable Reserved Instance for the predictable constrained baseline.
4. Use a Compute Savings Plan for the measured baseline commitment.
5. Use On-Demand while the short-lived or uncertain workload is characterized.

Correct Answer: B

 

Correct Answer

Answer B is correct because Spot uses spare EC2 capacity at discounted rates but can be interrupted, so it fits checkpointed or fault-tolerant work. This directly meets the decisive requirement: Spot Instances.

Incorrect Answers

Answer A is incorrect because Buying a commitment sized to peak demand risks paying for capacity when the burst disappears. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Spot Instances.

Answer C is incorrect because A stable, known EC2 usage pattern can trade flexibility for a larger discount where its family/Region constraints are acceptable. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Spot Instances.

Answer D is incorrect because Compute Savings Plans apply flexibly across eligible EC2, Fargate, and Lambda usage rather than binding the commitment to one instance family. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Spot Instances.

Answer E is incorrect because On-Demand avoids an up-front term commitment and is appropriate when future steady usage is not yet established. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Spot Instances.

 

Question 3

A new service has unknown traffic and may be shut down after three months. Which purchase model avoids a long commitment while demand is learned?

1. Evaluate an EC2 Instance Savings Plan or suitable Reserved Instance for the predictable constrained baseline.
2. Use On-Demand while the short-lived or uncertain workload is characterized.
3. Use a Compute Savings Plan for the measured baseline commitment.
4. Commit only the durable baseline and serve variable peaks with uncommitted or interruption-tolerant capacity as appropriate.
5. Use Spot Instances for the interruption-tolerant portion.

Correct Answer: B

 

Correct Answer

Answer B is correct because On-Demand avoids an up-front term commitment and is appropriate when future steady usage is not yet established. This directly meets the decisive requirement: On-Demand.

Incorrect Answers

Answer A is incorrect because A stable, known EC2 usage pattern can trade flexibility for a larger discount where its family/Region constraints are acceptable. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: On-Demand.

Answer C is incorrect because Compute Savings Plans apply flexibly across eligible EC2, Fargate, and Lambda usage rather than binding the commitment to one instance family. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: On-Demand.

Answer D is incorrect because Buying a commitment sized to peak demand risks paying for capacity when the burst disappears. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: On-Demand.

Answer E is incorrect because Spot uses spare EC2 capacity at discounted rates but can be interrupted, so it fits checkpointed or fault-tolerant work. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: On-Demand.

 

Question 4

A legacy licensed workload must run a known EC2 family in one Region for three years and usage is steady. Which discount model can be evaluated for the predictable baseline?

1. Evaluate an EC2 Instance Savings Plan or suitable Reserved Instance for the predictable constrained baseline.
2. Commit only the durable baseline and serve variable peaks with uncommitted or interruption-tolerant capacity as appropriate.
3. Use Spot Instances for the interruption-tolerant portion.
4. Use a Compute Savings Plan for the measured baseline commitment.
5. Use On-Demand while the short-lived or uncertain workload is characterized.

Correct Answer: A

 

Correct Answer

Answer A is correct because A stable, known EC2 usage pattern can trade flexibility for a larger discount where its family/Region constraints are acceptable. This directly meets the decisive requirement: EC2 Instance Savings Plan or suitable RI.

Incorrect Answers

Answer B is incorrect because Buying a commitment sized to peak demand risks paying for capacity when the burst disappears. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: EC2 Instance Savings Plan or suitable RI.

Answer C is incorrect because Spot uses spare EC2 capacity at discounted rates but can be interrupted, so it fits checkpointed or fault-tolerant work. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: EC2 Instance Savings Plan or suitable RI.

Answer D is incorrect because Compute Savings Plans apply flexibly across eligible EC2, Fargate, and Lambda usage rather than binding the commitment to one instance family. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: EC2 Instance Savings Plan or suitable RI.

Answer E is incorrect because On-Demand avoids an up-front term commitment and is appropriate when future steady usage is not yet established. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: EC2 Instance Savings Plan or suitable RI.

 

Question 5

A service has a steady baseline plus unpredictable peaks. Which approach best avoids committing to peak capacity?

1. Use Spot Instances for the interruption-tolerant portion.
2. Use a Compute Savings Plan for the measured baseline commitment.
3. Evaluate an EC2 Instance Savings Plan or suitable Reserved Instance for the predictable constrained baseline.
4. Use On-Demand while the short-lived or uncertain workload is characterized.
5. Commit only the durable baseline and serve variable peaks with uncommitted or interruption-tolerant capacity as appropriate.

Correct Answer: E

 

Correct Answer

Answer E is correct because Buying a commitment sized to peak demand risks paying for capacity when the burst disappears. This directly meets the decisive requirement: cover baseline with commitment and bursts On-Demand/eligible Spot.

Incorrect Answers

Answer A is incorrect because Spot uses spare EC2 capacity at discounted rates but can be interrupted, so it fits checkpointed or fault-tolerant work. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: cover baseline with commitment and bursts On-Demand/eligible Spot.

Answer B is incorrect because Compute Savings Plans apply flexibly across eligible EC2, Fargate, and Lambda usage rather than binding the commitment to one instance family. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: cover baseline with commitment and bursts On-Demand/eligible Spot.

Answer C is incorrect because A stable, known EC2 usage pattern can trade flexibility for a larger discount where its family/Region constraints are acceptable. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: cover baseline with commitment and bursts On-Demand/eligible Spot.

Answer D is incorrect because On-Demand avoids an up-front term commitment and is appropriate when future steady usage is not yet established. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: cover baseline with commitment and bursts On-Demand/eligible Spot.

 

Question 6

An m-family EC2 instance averages 4% CPU and 18% memory for 60 days with no forecast growth. Which first optimization should be evaluated?

1. Evaluate a compute-optimized EC2 family.
2. Evaluate an AWS Graviton-based instance family after compatibility and performance testing.
3. Use AWS Compute Optimizer recommendations and underlying utilization metrics.
4. Evaluate a memory-optimized EC2 family.
5. Downsize the instance after validating peak utilization, headroom, and future needs.

Correct Answer: E

 

Correct Answer

Answer E is correct because Sustained low utilization is evidence that a smaller instance may satisfy the workload at lower cost. This directly meets the decisive requirement: downsize after metric validation.

Incorrect Answers

Answer A is incorrect because Compute-optimized families allocate proportionally more CPU for workloads constrained primarily by compute rather than memory. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: downsize after metric validation.

Answer B is incorrect because Changing processor architecture can improve price/performance, but only when the workload and dependencies are compatible. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: downsize after metric validation.

Answer C is incorrect because Compute Optimizer analyzes historical resource configuration and metrics to produce rightsizing recommendations that still require workload judgment. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: downsize after metric validation.

Answer D is incorrect because Memory-optimized families are designed for workloads where memory, not CPU, is the limiting resource. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: downsize after metric validation.

 

Question 7

A compute job is CPU-bound at 95% but memory is mostly idle. Which instance-family direction should be evaluated?

1. Downsize the instance after validating peak utilization, headroom, and future needs.
2. Evaluate a compute-optimized EC2 family.
3. Evaluate a memory-optimized EC2 family.
4. Use AWS Compute Optimizer recommendations and underlying utilization metrics.
5. Evaluate an AWS Graviton-based instance family after compatibility and performance testing.

Correct Answer: B

 

Correct Answer

Answer B is correct because Compute-optimized families allocate proportionally more CPU for workloads constrained primarily by compute rather than memory. This directly meets the decisive requirement: compute-optimized family.

Incorrect Answers

Answer A is incorrect because Sustained low utilization is evidence that a smaller instance may satisfy the workload at lower cost. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: compute-optimized family.

Answer C is incorrect because Memory-optimized families are designed for workloads where memory, not CPU, is the limiting resource. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: compute-optimized family.

Answer D is incorrect because Compute Optimizer analyzes historical resource configuration and metrics to produce rightsizing recommendations that still require workload judgment. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: compute-optimized family.

Answer E is incorrect because Changing processor architecture can improve price/performance, but only when the workload and dependencies are compatible. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: compute-optimized family.

 

Question 8

A Java service uses 85% memory and 20% CPU. Which family direction is more appropriate than simply adding vCPUs?

1. Use AWS Compute Optimizer recommendations and underlying utilization metrics.
2. Downsize the instance after validating peak utilization, headroom, and future needs.
3. Evaluate a memory-optimized EC2 family.
4. Evaluate a compute-optimized EC2 family.
5. Evaluate an AWS Graviton-based instance family after compatibility and performance testing.

Correct Answer: C

 

Correct Answer

Answer C is correct because Memory-optimized families are designed for workloads where memory, not CPU, is the limiting resource. This directly meets the decisive requirement: memory-optimized family.

Incorrect Answers

Answer A is incorrect because Compute Optimizer analyzes historical resource configuration and metrics to produce rightsizing recommendations that still require workload judgment. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: memory-optimized family.

Answer B is incorrect because Sustained low utilization is evidence that a smaller instance may satisfy the workload at lower cost. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: memory-optimized family.

Answer D is incorrect because Compute-optimized families allocate proportionally more CPU for workloads constrained primarily by compute rather than memory. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: memory-optimized family.

Answer E is incorrect because Changing processor architecture can improve price/performance, but only when the workload and dependencies are compatible. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: memory-optimized family.

 

Question 9

A compatible stateless service currently runs x86 but testing shows the same performance on AWS Graviton. Which economic change can be considered?

1. Evaluate an AWS Graviton-based instance family after compatibility and performance testing.
2. Downsize the instance after validating peak utilization, headroom, and future needs.
3. Use AWS Compute Optimizer recommendations and underlying utilization metrics.
4. Evaluate a compute-optimized EC2 family.
5. Evaluate a memory-optimized EC2 family.

Correct Answer: A

 

Correct Answer

Answer A is correct because Changing processor architecture can improve price/performance, but only when the workload and dependencies are compatible. This directly meets the decisive requirement: move to suitable Graviton instances.

Incorrect Answers

Answer B is incorrect because Sustained low utilization is evidence that a smaller instance may satisfy the workload at lower cost. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: move to suitable Graviton instances.

Answer C is incorrect because Compute Optimizer analyzes historical resource configuration and metrics to produce rightsizing recommendations that still require workload judgment. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: move to suitable Graviton instances.

Answer D is incorrect because Compute-optimized families allocate proportionally more CPU for workloads constrained primarily by compute rather than memory. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: move to suitable Graviton instances.

Answer E is incorrect because Memory-optimized families are designed for workloads where memory, not CPU, is the limiting resource. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: move to suitable Graviton instances.

 

Question 10

A team wants evidence-based EC2 rightsizing recommendations using historical utilization metrics. Which AWS service should be consulted?

1. Downsize the instance after validating peak utilization, headroom, and future needs.
2. Evaluate a compute-optimized EC2 family.
3. Evaluate a memory-optimized EC2 family.
4. Use AWS Compute Optimizer recommendations and underlying utilization metrics.
5. Evaluate an AWS Graviton-based instance family after compatibility and performance testing.

Correct Answer: D

 

Correct Answer

Answer D is correct because Compute Optimizer analyzes historical resource configuration and metrics to produce rightsizing recommendations that still require workload judgment. This directly meets the decisive requirement: Compute Optimizer.

Incorrect Answers

Answer A is incorrect because Sustained low utilization is evidence that a smaller instance may satisfy the workload at lower cost. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Compute Optimizer.

Answer B is incorrect because Compute-optimized families allocate proportionally more CPU for workloads constrained primarily by compute rather than memory. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Compute Optimizer.

Answer C is incorrect because Memory-optimized families are designed for workloads where memory, not CPU, is the limiting resource. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Compute Optimizer.

Answer E is incorrect because Changing processor architecture can improve price/performance, but only when the workload and dependencies are compatible. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Compute Optimizer.

 

Question 11

Development EC2 instances are unused every night and weekend and have no after-hours availability requirement. Which action directly reduces idle compute cost?

1. Terminate orphaned instances after ownership and dependency checks.
2. Use EC2 hibernation where supported and operationally appropriate.
3. Schedule nonproduction instances to stop and start around required working hours.
4. Use queue-driven or scheduled automation that starts capacity for work and scales it down when the backlog clears.
5. Lower or schedule Auto Scaling minimum capacity to the proven demand floor while preserving required availability.

Correct Answer: C

 

Correct Answer

Answer C is correct because EC2 instance compute charges stop while stopped, although attached storage and other resources can continue to incur cost. This directly meets the decisive requirement: scheduled stop/start.

Incorrect Answers

Answer A is incorrect because Completed workloads that no longer provide service should not continue consuming compute merely because cleanup was missed. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: scheduled stop/start.

Answer B is incorrect because Hibernation preserves in-memory state to the root volume while the instance is stopped, trading storage for reduced running time. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: scheduled stop/start.

Answer D is incorrect because Aligning worker lifetime with actual backlog eliminates manual idle periods after processing completes. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: scheduled stop/start.

Answer E is incorrect because A minimum capacity above actual need forces idle instances to run regardless of scaling signals. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: scheduled stop/start.

 

Question 12

A stateful workstation needs RAM contents preserved overnight but can be unavailable during that time and supports the feature. Which EC2 action may reduce running cost while preserving memory state?

1. Schedule nonproduction instances to stop and start around required working hours.
2. Use queue-driven or scheduled automation that starts capacity for work and scales it down when the backlog clears.
3. Terminate orphaned instances after ownership and dependency checks.
4. Use EC2 hibernation where supported and operationally appropriate.
5. Lower or schedule Auto Scaling minimum capacity to the proven demand floor while preserving required availability.

Correct Answer: D

 

Correct Answer

Answer D is correct because Hibernation preserves in-memory state to the root volume while the instance is stopped, trading storage for reduced running time. This directly meets the decisive requirement: hibernate supported instance.

Incorrect Answers

Answer A is incorrect because EC2 instance compute charges stop while stopped, although attached storage and other resources can continue to incur cost. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: hibernate supported instance.

Answer B is incorrect because Aligning worker lifetime with actual backlog eliminates manual idle periods after processing completes. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: hibernate supported instance.

Answer C is incorrect because Completed workloads that no longer provide service should not continue consuming compute merely because cleanup was missed. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: hibernate supported instance.

Answer E is incorrect because A minimum capacity above actual need forces idle instances to run regardless of scaling signals. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: hibernate supported instance.

 

Question 13

An Auto Scaling group has minimum capacity 10 although observed overnight demand needs only 2 healthy instances. Which setting should be reviewed?

1. Terminate orphaned instances after ownership and dependency checks.
2. Use queue-driven or scheduled automation that starts capacity for work and scales it down when the backlog clears.
3. Schedule nonproduction instances to stop and start around required working hours.
4. Lower or schedule Auto Scaling minimum capacity to the proven demand floor while preserving required availability.
5. Use EC2 hibernation where supported and operationally appropriate.

Correct Answer: D

 

Correct Answer

Answer D is correct because A minimum capacity above actual need forces idle instances to run regardless of scaling signals. This directly meets the decisive requirement: reduce minimum or scheduled capacity within SLA.

Incorrect Answers

Answer A is incorrect because Completed workloads that no longer provide service should not continue consuming compute merely because cleanup was missed. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: reduce minimum or scheduled capacity within SLA.

Answer B is incorrect because Aligning worker lifetime with actual backlog eliminates manual idle periods after processing completes. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: reduce minimum or scheduled capacity within SLA.

Answer C is incorrect because EC2 instance compute charges stop while stopped, although attached storage and other resources can continue to incur cost. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: reduce minimum or scheduled capacity within SLA.

Answer E is incorrect because Hibernation preserves in-memory state to the root volume while the instance is stopped, trading storage for reduced running time. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: reduce minimum or scheduled capacity within SLA.

 

Question 14

A one-time analytics cluster finished last month but instances still run because no owner removed them. Which governance action is most direct?

1. Use queue-driven or scheduled automation that starts capacity for work and scales it down when the backlog clears.
2. Terminate orphaned instances after ownership and dependency checks.
3. Lower or schedule Auto Scaling minimum capacity to the proven demand floor while preserving required availability.
4. Use EC2 hibernation where supported and operationally appropriate.
5. Schedule nonproduction instances to stop and start around required working hours.

Correct Answer: B

 

Correct Answer

Answer B is correct because Completed workloads that no longer provide service should not continue consuming compute merely because cleanup was missed. This directly meets the decisive requirement: identify owner and terminate unused compute.

Incorrect Answers

Answer A is incorrect because Aligning worker lifetime with actual backlog eliminates manual idle periods after processing completes. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: identify owner and terminate unused compute.

Answer C is incorrect because A minimum capacity above actual need forces idle instances to run regardless of scaling signals. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: identify owner and terminate unused compute.

Answer D is incorrect because Hibernation preserves in-memory state to the root volume while the instance is stopped, trading storage for reduced running time. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: identify owner and terminate unused compute.

Answer E is incorrect because EC2 instance compute charges stop while stopped, although attached storage and other resources can continue to incur cost. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: identify owner and terminate unused compute.

 

Question 15

A batch fleet is launched manually every morning and is often left running after the queue empties. Which design better aligns runtime with work?

1. Use EC2 hibernation where supported and operationally appropriate.
2. Use queue-driven or scheduled automation that starts capacity for work and scales it down when the backlog clears.
3. Schedule nonproduction instances to stop and start around required working hours.
4. Terminate orphaned instances after ownership and dependency checks.
5. Lower or schedule Auto Scaling minimum capacity to the proven demand floor while preserving required availability.

Correct Answer: B

 

Correct Answer

Answer B is correct because Aligning worker lifetime with actual backlog eliminates manual idle periods after processing completes. This directly meets the decisive requirement: automated scale-to-work / terminate workers.

Incorrect Answers

Answer A is incorrect because Hibernation preserves in-memory state to the root volume while the instance is stopped, trading storage for reduced running time. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: automated scale-to-work / terminate workers.

Answer C is incorrect because EC2 instance compute charges stop while stopped, although attached storage and other resources can continue to incur cost. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: automated scale-to-work / terminate workers.

Answer D is incorrect because Completed workloads that no longer provide service should not continue consuming compute merely because cleanup was missed. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: automated scale-to-work / terminate workers.

Answer E is incorrect because A minimum capacity above actual need forces idle instances to run regardless of scaling signals. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: automated scale-to-work / terminate workers.

 

Question 16

A function runs for two seconds a few hundred times per day after S3 uploads. Which execution model most directly avoids paying for idle servers?

2. Use Amazon EC2.
3. Use AWS Lambda.
4. Use AWS Fargate for the container service.
5. Use a serverless function model such as Lambda when its service limits fit.
6. Use eligible Spot capacity for the interruption-tolerant compute pool.

Correct Answer: B

 

Correct Answer

Answer B is correct because Lambda bills around requests and execution duration and eliminates permanently provisioned server capacity for short event-driven work. This directly meets the decisive requirement: Lambda.

Incorrect Answers

Answer A is incorrect because EC2 provides operating-system and instance-level control needed by workloads that cannot fit managed serverless/container abstractions. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Lambda.

Answer C is incorrect because Fargate runs containers without managing EC2 hosts and charges for task resources while tasks run. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Lambda.

Answer D is incorrect because When the workload is mostly idle, per-invocation execution avoids paying continuously for idle compute capacity. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Lambda.

Answer E is incorrect because Spot can materially reduce cost when the workload can tolerate capacity reclamation and restart elsewhere. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Lambda.

 

Question 17

A containerized API runs continuously with predictable baseline utilization but the team does not want to manage EC2 hosts. Which managed compute option fits?

1. Use AWS Fargate for the container service.
2. Use AWS Lambda.
3. Use eligible Spot capacity for the interruption-tolerant compute pool.
4. Use Amazon EC2.
5. Use a serverless function model such as Lambda when its service limits fit.

Correct Answer: A

 

Correct Answer

Answer A is correct because Fargate runs containers without managing EC2 hosts and charges for task resources while tasks run. This directly meets the decisive requirement: Fargate.

Incorrect Answers

Answer B is incorrect because Lambda bills around requests and execution duration and eliminates permanently provisioned server capacity for short event-driven work. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Fargate.

Answer C is incorrect because Spot can materially reduce cost when the workload can tolerate capacity reclamation and restart elsewhere. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Fargate.

Answer D is incorrect because EC2 provides operating-system and instance-level control needed by workloads that cannot fit managed serverless/container abstractions. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Fargate.

Answer E is incorrect because When the workload is mostly idle, per-invocation execution avoids paying continuously for idle compute capacity. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Fargate.

 

Question 18

A licensed appliance requires kernel modules and full operating-system control. Which compute model is necessary?

1. Use eligible Spot capacity for the interruption-tolerant compute pool.
2. Use AWS Fargate for the container service.
3. Use a serverless function model such as Lambda when its service limits fit.
4. Use Amazon EC2.
5. Use AWS Lambda.

Correct Answer: D

 

Correct Answer

Answer D is correct because EC2 provides operating-system and instance-level control needed by workloads that cannot fit managed serverless/container abstractions. This directly meets the decisive requirement: EC2.

Incorrect Answers

Answer A is incorrect because Spot can materially reduce cost when the workload can tolerate capacity reclamation and restart elsewhere. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: EC2.

Answer B is incorrect because Fargate runs containers without managing EC2 hosts and charges for task resources while tasks run. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: EC2.

Answer C is incorrect because When the workload is mostly idle, per-invocation execution avoids paying continuously for idle compute capacity. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: EC2.

Answer E is incorrect because Lambda bills around requests and execution duration and eliminates permanently provisioned server capacity for short event-driven work. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: EC2.

 

Question 19

A fault-tolerant container batch job can restart tasks and is not latency sensitive. Which capacity option can reduce EC2-backed container cost when architecture permits?

1. Use eligible Spot capacity for the interruption-tolerant compute pool.
2. Use AWS Fargate for the container service.
3. Use Amazon EC2.
4. Use a serverless function model such as Lambda when its service limits fit.
5. Use AWS Lambda.

Correct Answer: A

 

Correct Answer

Answer A is correct because Spot can materially reduce cost when the workload can tolerate capacity reclamation and restart elsewhere. This directly meets the decisive requirement: Spot capacity.

Incorrect Answers

Answer B is incorrect because Fargate runs containers without managing EC2 hosts and charges for task resources while tasks run. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Spot capacity.

Answer C is incorrect because EC2 provides operating-system and instance-level control needed by workloads that cannot fit managed serverless/container abstractions. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Spot capacity.

Answer D is incorrect because When the workload is mostly idle, per-invocation execution avoids paying continuously for idle compute capacity. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Spot capacity.

Answer E is incorrect because Lambda bills around requests and execution duration and eliminates permanently provisioned server capacity for short event-driven work. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Spot capacity.

 

Question 20

A tiny API has long idle periods and bursty requests with sub-15-minute execution. Which economic characteristic favors serverless functions?

1. Use AWS Lambda.
2. Use Amazon EC2.
3. Use AWS Fargate for the container service.
4. Use eligible Spot capacity for the interruption-tolerant compute pool.
5. Use a serverless function model such as Lambda when its service limits fit.

Correct Answer: E

 

Correct Answer

Answer E is correct because When the workload is mostly idle, per-invocation execution avoids paying continuously for idle compute capacity. This directly meets the decisive requirement: pay per execution rather than idle capacity.

Incorrect Answers

Answer A is incorrect because Lambda bills around requests and execution duration and eliminates permanently provisioned server capacity for short event-driven work. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: pay per execution rather than idle capacity.

Answer B is incorrect because EC2 provides operating-system and instance-level control needed by workloads that cannot fit managed serverless/container abstractions. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: pay per execution rather than idle capacity.

Answer C is incorrect because Fargate runs containers without managing EC2 hosts and charges for task resources while tasks run. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: pay per execution rather than idle capacity.

Answer D is incorrect because Spot can materially reduce cost when the workload can tolerate capacity reclamation and restart elsewhere. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: pay per execution rather than idle capacity.

 

Question 21

Ten small HTTP services each use a separate ALB only for host routing, and one shared ALB can meet isolation and quota requirements. Which consolidation may reduce fixed load-balancer overhead?

1. Use a Gateway Load Balancer.
2. Right-size and autoscale the backend target fleets.
3. Use an Application Load Balancer because the workload needs its Layer 7 routing and WebSocket capabilities.
4. Consolidate compatible HTTP services behind shared ALB listeners and host/path rules when security and limits permit.
5. Use a Network Load Balancer for the Layer 4 TCP requirement.

Correct Answer: D

 

Correct Answer

Answer D is correct because A shared Layer 7 entry point can remove redundant fixed balancer resources while maintaining logical target separation. This directly meets the decisive requirement: consolidate with host/path rules.

Incorrect Answers

Answer A is incorrect because GWLB is designed to deploy and scale fleets of virtual network appliances and steer traffic transparently through them. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: consolidate with host/path rules.

Answer B is incorrect because When target instances are mostly idle, reducing backend overprovisioning provides a larger direct compute saving than changing the already-shared load balancer. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: consolidate with host/path rules.

Answer C is incorrect because Cost optimization must preserve required functionality; ALB features are justified when the application explicitly depends on them. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: consolidate with host/path rules.

Answer E is incorrect because NLB handles TCP/UDP/TLS traffic without requiring Layer 7 HTTP routing features. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: consolidate with host/path rules.

 

Question 22

A custom TCP service does not need Layer 7 features. Which load balancer avoids paying for application-routing features the protocol cannot use?

1. Use a Gateway Load Balancer.
2. Use an Application Load Balancer because the workload needs its Layer 7 routing and WebSocket capabilities.
3. Consolidate compatible HTTP services behind shared ALB listeners and host/path rules when security and limits permit.
4. Right-size and autoscale the backend target fleets.
5. Use a Network Load Balancer for the Layer 4 TCP requirement.

Correct Answer: E

 

Correct Answer

Answer E is correct because NLB handles TCP/UDP/TLS traffic without requiring Layer 7 HTTP routing features. This directly meets the decisive requirement: Network Load Balancer.

Incorrect Answers

Answer A is incorrect because GWLB is designed to deploy and scale fleets of virtual network appliances and steer traffic transparently through them. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Network Load Balancer.

Answer B is incorrect because Cost optimization must preserve required functionality; ALB features are justified when the application explicitly depends on them. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Network Load Balancer.

Answer C is incorrect because A shared Layer 7 entry point can remove redundant fixed balancer resources while maintaining logical target separation. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Network Load Balancer.

Answer D is incorrect because When target instances are mostly idle, reducing backend overprovisioning provides a larger direct compute saving than changing the already-shared load balancer. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Network Load Balancer.

 

Question 23

A security architecture must scale a fleet of transparent virtual firewalls. Which load-balancer type is purpose-built for this appliance pattern?

1. Use a Gateway Load Balancer.
2. Use an Application Load Balancer because the workload needs its Layer 7 routing and WebSocket capabilities.
3. Consolidate compatible HTTP services behind shared ALB listeners and host/path rules when security and limits permit.
4. Right-size and autoscale the backend target fleets.
5. Use a Network Load Balancer for the Layer 4 TCP requirement.

Correct Answer: A

 

Correct Answer

Answer A is correct because GWLB is designed to deploy and scale fleets of virtual network appliances and steer traffic transparently through them. This directly meets the decisive requirement: Gateway Load Balancer.

Incorrect Answers

Answer B is incorrect because Cost optimization must preserve required functionality; ALB features are justified when the application explicitly depends on them. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Gateway Load Balancer.

Answer C is incorrect because A shared Layer 7 entry point can remove redundant fixed balancer resources while maintaining logical target separation. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Gateway Load Balancer.

Answer D is incorrect because When target instances are mostly idle, reducing backend overprovisioning provides a larger direct compute saving than changing the already-shared load balancer. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Gateway Load Balancer.

Answer E is incorrect because NLB handles TCP/UDP/TLS traffic without requiring Layer 7 HTTP routing features. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Gateway Load Balancer.

 

Question 24

An ALB has 50 mostly idle targets because each microservice reserves five dedicated instances even at zero load. Which cost area should be optimized first?

1. Use an Application Load Balancer because the workload needs its Layer 7 routing and WebSocket capabilities.
2. Right-size and autoscale the backend target fleets.
3. Use a Gateway Load Balancer.
4. Use a Network Load Balancer for the Layer 4 TCP requirement.
5. Consolidate compatible HTTP services behind shared ALB listeners and host/path rules when security and limits permit.

Correct Answer: B

 

Correct Answer

Answer B is correct because When target instances are mostly idle, reducing backend overprovisioning provides a larger direct compute saving than changing the already-shared load balancer. This directly meets the decisive requirement: backend target scaling.

Incorrect Answers

Answer A is incorrect because Cost optimization must preserve required functionality; ALB features are justified when the application explicitly depends on them. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: backend target scaling.

Answer C is incorrect because GWLB is designed to deploy and scale fleets of virtual network appliances and steer traffic transparently through them. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: backend target scaling.

Answer D is incorrect because NLB handles TCP/UDP/TLS traffic without requiring Layer 7 HTTP routing features. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: backend target scaling.

Answer E is incorrect because A shared Layer 7 entry point can remove redundant fixed balancer resources while maintaining logical target separation. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: backend target scaling.

 

Question 25

A workload requires HTTP host/path routing and WebSocket support. Which balancer feature set justifies using an ALB rather than selecting by lower headline cost alone?

1. Right-size and autoscale the backend target fleets.
2. Use an Application Load Balancer because the workload needs its Layer 7 routing and WebSocket capabilities.
3. Consolidate compatible HTTP services behind shared ALB listeners and host/path rules when security and limits permit.
4. Use a Network Load Balancer for the Layer 4 TCP requirement.
5. Use a Gateway Load Balancer.

Correct Answer: B

 

Correct Answer

Answer B is correct because Cost optimization must preserve required functionality; ALB features are justified when the application explicitly depends on them. This directly meets the decisive requirement: Application Load Balancer.

Incorrect Answers

Answer A is incorrect because When target instances are mostly idle, reducing backend overprovisioning provides a larger direct compute saving than changing the already-shared load balancer. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Application Load Balancer.

Answer C is incorrect because A shared Layer 7 entry point can remove redundant fixed balancer resources while maintaining logical target separation. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Application Load Balancer.

Answer D is incorrect because NLB handles TCP/UDP/TLS traffic without requiring Layer 7 HTTP routing features. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Application Load Balancer.

Answer E is incorrect because GWLB is designed to deploy and scale fleets of virtual network appliances and steer traffic transparently through them. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Application Load Balancer.

 

Question 26

A platform team needs to compare monthly EC2 spend across member accounts. Which Cost Management view is a direct starting point?

1. Use Cost Explorer grouped or filtered by linked account and service.
2. Compare the pre- and post-release periods in Cost Explorer and group by relevant service, usage type, account, or tag.
3. Use AWS Cost and Usage Reports.
4. Use AWS Budgets with forecast notifications.
5. Adopt consistent cost allocation tags and activate them for billing.

Correct Answer: A

 

Correct Answer

Answer A is correct because Cost Explorer is designed for interactive cost/usage analysis across accounts, services, tags, and time periods. This directly meets the decisive requirement: Cost Explorer grouping.

Incorrect Answers

Answer B is incorrect because Time-based cost analysis identifies which compute usage changed rather than guessing from the release alone. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Cost Explorer grouping.

Answer C is incorrect because CUR provides detailed line items suitable for downstream SQL/warehouse analysis and can include resource identifiers and tags. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Cost Explorer grouping.

Answer D is incorrect because Budgets can alert when actual or forecast costs/usage cross configured thresholds. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Cost Explorer grouping.

Answer E is incorrect because Resource tags only become useful allocation dimensions when governance ensures coverage and the tags are activated in cost reporting. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Cost Explorer grouping.

 

Question 27

Finance needs per-resource compute cost data with usage types for a custom data warehouse. Which billing data source is appropriate?

1. Compare the pre- and post-release periods in Cost Explorer and group by relevant service, usage type, account, or tag.
2. Adopt consistent cost allocation tags and activate them for billing.
3. Use Cost Explorer grouped or filtered by linked account and service.
4. Use AWS Cost and Usage Reports.
5. Use AWS Budgets with forecast notifications.

Correct Answer: D

 

Correct Answer

Answer D is correct because CUR provides detailed line items suitable for downstream SQL/warehouse analysis and can include resource identifiers and tags. This directly meets the decisive requirement: Cost and Usage Report.

Incorrect Answers

Answer A is incorrect because Time-based cost analysis identifies which compute usage changed rather than guessing from the release alone. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Cost and Usage Report.

Answer B is incorrect because Resource tags only become useful allocation dimensions when governance ensures coverage and the tags are activated in cost reporting. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Cost and Usage Report.

Answer C is incorrect because Cost Explorer is designed for interactive cost/usage analysis across accounts, services, tags, and time periods. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Cost and Usage Report.

Answer E is incorrect because Budgets can alert when actual or forecast costs/usage cross configured thresholds. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Cost and Usage Report.

 

Question 28

A service owner wants notification if compute spend is forecast to exceed the quarterly target. Which feature is appropriate?

1. Compare the pre- and post-release periods in Cost Explorer and group by relevant service, usage type, account, or tag.
2. Use Cost Explorer grouped or filtered by linked account and service.
3. Use AWS Budgets with forecast notifications.
4. Adopt consistent cost allocation tags and activate them for billing.
5. Use AWS Cost and Usage Reports.

Correct Answer: C

 

Correct Answer

Answer C is correct because Budgets can alert when actual or forecast costs/usage cross configured thresholds. This directly meets the decisive requirement: AWS Budgets.

Incorrect Answers

Answer A is incorrect because Time-based cost analysis identifies which compute usage changed rather than guessing from the release alone. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: AWS Budgets.

Answer B is incorrect because Cost Explorer is designed for interactive cost/usage analysis across accounts, services, tags, and time periods. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: AWS Budgets.

Answer D is incorrect because Resource tags only become useful allocation dimensions when governance ensures coverage and the tags are activated in cost reporting. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: AWS Budgets.

Answer E is incorrect because CUR provides detailed line items suitable for downstream SQL/warehouse analysis and can include resource identifiers and tags. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: AWS Budgets.

 

Question 29

A newly deployed release doubled EC2 spend. Which analysis should be performed first?

1. Use AWS Budgets with forecast notifications.
2. Adopt consistent cost allocation tags and activate them for billing.
3. Compare the pre- and post-release periods in Cost Explorer and group by relevant service, usage type, account, or tag.
4. Use AWS Cost and Usage Reports.
5. Use Cost Explorer grouped or filtered by linked account and service.

Correct Answer: C

 

Correct Answer

Answer C is correct because Time-based cost analysis identifies which compute usage changed rather than guessing from the release alone. This directly meets the decisive requirement: Cost Explorer time/filter comparison.

Incorrect Answers

Answer A is incorrect because Budgets can alert when actual or forecast costs/usage cross configured thresholds. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Cost Explorer time/filter comparison.

Answer B is incorrect because Resource tags only become useful allocation dimensions when governance ensures coverage and the tags are activated in cost reporting. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Cost Explorer time/filter comparison.

Answer D is incorrect because CUR provides detailed line items suitable for downstream SQL/warehouse analysis and can include resource identifiers and tags. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Cost Explorer time/filter comparison.

Answer E is incorrect because Cost Explorer is designed for interactive cost/usage analysis across accounts, services, tags, and time periods. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: Cost Explorer time/filter comparison.

 

Question 30

Several teams share one account and finance cannot attribute EC2 resources to products. Which governance improvement enables future allocation?

1. Adopt consistent cost allocation tags and activate them for billing.
2. Use AWS Budgets with forecast notifications.
3. Compare the pre- and post-release periods in Cost Explorer and group by relevant service, usage type, account, or tag.
4. Use AWS Cost and Usage Reports.
5. Use Cost Explorer grouped or filtered by linked account and service.

Correct Answer: A

 

Correct Answer

Answer A is correct because Resource tags only become useful allocation dimensions when governance ensures coverage and the tags are activated in cost reporting. This directly meets the decisive requirement: consistent activated cost allocation tags.

Incorrect Answers

Answer B is incorrect because Budgets can alert when actual or forecast costs/usage cross configured thresholds. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: consistent activated cost allocation tags.

Answer C is incorrect because Time-based cost analysis identifies which compute usage changed rather than guessing from the release alone. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: consistent activated cost allocation tags.

Answer D is incorrect because CUR provides detailed line items suitable for downstream SQL/warehouse analysis and can include resource identifiers and tags. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: consistent activated cost allocation tags.

Answer E is incorrect because Cost Explorer is designed for interactive cost/usage analysis across accounts, services, tags, and time periods. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: consistent activated cost allocation tags.

 

Question 31

A customer-facing stateful database cannot tolerate sudden capacity loss. Which purchasing choice should NOT be used as the only capacity source solely to save money?

1. Maintain an On-Demand or committed baseline and diversify Spot for the flexible remainder.
2. Base the commitment on the durable future baseline rather than blindly accepting a recommendation derived from higher historical usage.
3. Retain justified capacity headroom or improve scaling responsiveness so the service-level objective is met.
4. Preserve enough multi-AZ capacity headroom to meet the documented failure objective.
5. Use reliable non-interruptible capacity for the critical baseline instead of depending only on Spot.

Correct Answer: E

 

Correct Answer

Answer E is correct because Spot can be interrupted and AWS recommends it for interruption-tolerant workloads rather than as the sole source for workloads that cannot tolerate loss. This directly meets the decisive requirement: do not rely solely on Spot.

Incorrect Answers

Answer A is incorrect because A mixed strategy captures Spot savings while preserving service when Spot capacity is reclaimed or temporarily unavailable. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: do not rely solely on Spot.

Answer B is incorrect because Savings commitments are difficult to unwind, so expected future usage must be considered before purchase. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: do not rely solely on Spot.

Answer C is incorrect because A cheaper fleet that consistently fails during bursts is not a valid optimization when performance and availability requirements are explicit. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: do not rely solely on Spot.

Answer D is incorrect because Eliminating all spare capacity can reduce cost but violates the stated requirement to survive loss of an Availability Zone. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: do not rely solely on Spot.

 

Question 32

A stateless web tier can use Spot but must sustain traffic during Spot shortages. Which cost-aware design is appropriate?

1. Retain justified capacity headroom or improve scaling responsiveness so the service-level objective is met.
2. Base the commitment on the durable future baseline rather than blindly accepting a recommendation derived from higher historical usage.
3. Maintain an On-Demand or committed baseline and diversify Spot for the flexible remainder.
4. Use reliable non-interruptible capacity for the critical baseline instead of depending only on Spot.
5. Preserve enough multi-AZ capacity headroom to meet the documented failure objective.

Correct Answer: C

 

Correct Answer

Answer C is correct because A mixed strategy captures Spot savings while preserving service when Spot capacity is reclaimed or temporarily unavailable. This directly meets the decisive requirement: mixed On-Demand baseline plus diversified Spot.

Incorrect Answers

Answer A is incorrect because A cheaper fleet that consistently fails during bursts is not a valid optimization when performance and availability requirements are explicit. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: mixed On-Demand baseline plus diversified Spot.

Answer B is incorrect because Savings commitments are difficult to unwind, so expected future usage must be considered before purchase. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: mixed On-Demand baseline plus diversified Spot.

Answer D is incorrect because Spot can be interrupted and AWS recommends it for interruption-tolerant workloads rather than as the sole source for workloads that cannot tolerate loss. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: mixed On-Demand baseline plus diversified Spot.

Answer E is incorrect because Eliminating all spare capacity can reduce cost but violates the stated requirement to survive loss of an Availability Zone. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: mixed On-Demand baseline plus diversified Spot.

 

Question 33

A production fleet is right-sized so tightly that one AZ failure leaves no capacity headroom. What should the architect do?

1. Maintain an On-Demand or committed baseline and diversify Spot for the flexible remainder.
2. Retain justified capacity headroom or improve scaling responsiveness so the service-level objective is met.
3. Base the commitment on the durable future baseline rather than blindly accepting a recommendation derived from higher historical usage.
4. Use reliable non-interruptible capacity for the critical baseline instead of depending only on Spot.
5. Preserve enough multi-AZ capacity headroom to meet the documented failure objective.

Correct Answer: E

 

Correct Answer

Answer E is correct because Eliminating all spare capacity can reduce cost but violates the stated requirement to survive loss of an Availability Zone. This directly meets the decisive requirement: retain failure capacity even if average utilization rises.

Incorrect Answers

Answer A is incorrect because A mixed strategy captures Spot savings while preserving service when Spot capacity is reclaimed or temporarily unavailable. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: retain failure capacity even if average utilization rises.

Answer B is incorrect because A cheaper fleet that consistently fails during bursts is not a valid optimization when performance and availability requirements are explicit. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: retain failure capacity even if average utilization rises.

Answer C is incorrect because Savings commitments are difficult to unwind, so expected future usage must be considered before purchase. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: retain failure capacity even if average utilization rises.

Answer D is incorrect because Spot can be interrupted and AWS recommends it for interruption-tolerant workloads rather than as the sole source for workloads that cannot tolerate loss. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: retain failure capacity even if average utilization rises.

 

Question 34

A Savings Plan recommendation covers 100% of last month usage, but demand is seasonal and expected to fall sharply. Which action is prudent?

1. Maintain an On-Demand or committed baseline and diversify Spot for the flexible remainder.
2. Base the commitment on the durable future baseline rather than blindly accepting a recommendation derived from higher historical usage.
3. Preserve enough multi-AZ capacity headroom to meet the documented failure objective.
4. Retain justified capacity headroom or improve scaling responsiveness so the service-level objective is met.
5. Use reliable non-interruptible capacity for the critical baseline instead of depending only on Spot.

Correct Answer: B

 

Correct Answer

Answer B is correct because Savings commitments are difficult to unwind, so expected future usage must be considered before purchase. This directly meets the decisive requirement: commit only durable expected baseline.

Incorrect Answers

Answer A is incorrect because A mixed strategy captures Spot savings while preserving service when Spot capacity is reclaimed or temporarily unavailable. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: commit only durable expected baseline.

Answer C is incorrect because Eliminating all spare capacity can reduce cost but violates the stated requirement to survive loss of an Availability Zone. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: commit only durable expected baseline.

Answer D is incorrect because A cheaper fleet that consistently fails during bursts is not a valid optimization when performance and availability requirements are explicit. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: commit only durable expected baseline.

Answer E is incorrect because Spot can be interrupted and AWS recommends it for interruption-tolerant workloads rather than as the sole source for workloads that cannot tolerate loss. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: commit only durable expected baseline.

 

Question 35

A scale-out policy adds capacity only after long warm-up and users see errors during bursts. Which statement is correct about cost optimization?

1. Base the commitment on the durable future baseline rather than blindly accepting a recommendation derived from higher historical usage.
2. Use reliable non-interruptible capacity for the critical baseline instead of depending only on Spot.
3. Retain justified capacity headroom or improve scaling responsiveness so the service-level objective is met.
4. Preserve enough multi-AZ capacity headroom to meet the documented failure objective.
5. Maintain an On-Demand or committed baseline and diversify Spot for the flexible remainder.

Correct Answer: C

 

Correct Answer

Answer C is correct because A cheaper fleet that consistently fails during bursts is not a valid optimization when performance and availability requirements are explicit. This directly meets the decisive requirement: keep justified headroom or faster scaling even if utilization is lower.

Incorrect Answers

Answer A is incorrect because Savings commitments are difficult to unwind, so expected future usage must be considered before purchase. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: keep justified headroom or faster scaling even if utilization is lower.

Answer B is incorrect because Spot can be interrupted and AWS recommends it for interruption-tolerant workloads rather than as the sole source for workloads that cannot tolerate loss. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: keep justified headroom or faster scaling even if utilization is lower.

Answer D is incorrect because Eliminating all spare capacity can reduce cost but violates the stated requirement to survive loss of an Availability Zone. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: keep justified headroom or faster scaling even if utilization is lower.

Answer E is incorrect because A mixed strategy captures Spot savings while preserving service when Spot capacity is reclaimed or temporarily unavailable. It can be useful in a different cost or architecture situation, but it does not meet the decisive requirement here: keep justified headroom or faster scaling even if utilization is lower.