SAA-C03 - Design Cost-Optimized Architectures (20% of the exam) - Section 4.3

Select cost-effective compute pricing models including Spot Instances, Reserved Instances and Savings Plans.

Compare On-Demand, Reserved Instances, Savings Plans, and Spot Instances on commitment term, discount depth, and workload suitability. Choose Spot Instances for fault-tolerant, interruption-resilient workloads such as batch processing to achieve the deepest discounts, Reserved Instances or Compute Savings Plans for steady-state production workloads to commit capacity and reduce hourly cost, and On-Demand for short-lived or unpredictable workloads.

Spot InstancesReserved InstancesSavings PlansOn-Demand pricing

Practice question for this objective

Free sampleDesign Cost-Optimized Architectureshard

A media company processes a nightly video transcoding queue on a fleet of EC2 instances. Each job is short, fully stateless, and checkpoints its progress to Amazon S3, so any instance that disappears mid-job simply has its work resumed by another. The fleet can flex between many instance types and runs only when the queue is non-empty. The architect wants the lowest possible compute cost without committing to any fixed term. Which pricing model best fits this workload?

AStandard Reserved Instances purchased for a one-year term against the most common instance type, accepting the term commitment to lock in the reservation discount on the fleet.
BOn-Demand Instances run continuously so the queue is always drained immediately, paying the full hourly rate in exchange for never being interrupted by a capacity reclaim.
CA one-year Compute Savings Plan sized to the average nightly throughput, committing to a steady hourly spend to discount the transcoding fleet across instance families.
DSpot Instances drawn from several instance pools, letting the interruption-tolerant fleet reclaim and resume checkpointed jobs whenever capacity is reclaimed by AWS at short notice. Correct

Choose Spot Instances for stateless, interruption-tolerant, instance-flexible workloads that checkpoint progress and need the deepest discount with no term commitment. Spot Instances sell spare EC2 capacity at the largest discount off On-Demand, but AWS can reclaim them with two minutes of notice. A transcoding fleet that is stateless, checkpoints to S3, can run across many pools, and resumes interrupted work tolerates that reclamation with no data loss, so Spot captures the saving while term-based models would impose an unwanted commitment.

Why A is wrong: Tempting because Reserved Instances cut cost on steady usage, but they require a one-year or three-year commitment and suit always-on baselines, not an interruption-tolerant fleet that runs only when work exists.

Why B is wrong: Tempting because On-Demand needs no commitment, but it charges the full rate and forgoes the deep discount this interruptible, checkpointing fleet is perfectly positioned to capture with Spot.

Why C is wrong: Tempting because Savings Plans discount flexible compute, but they demand a one-year or three-year hourly spend commitment that does not match a bursty queue, and the stem rules out any fixed term.

Why D is correct: Spot offers the deepest discount off On-Demand and is built for stateless, fault-tolerant, flexible workloads that checkpoint progress, so interruptions cost nothing beyond a resumed job.

See more SAA-C03 practice questions, answers explained.

Exam traps in Design Cost-Optimized Architectures

Answers that look right on this material and are not. Each one is a distractor from a different question in the SAA-C03 bank for this domain.

A three-year Compute Savings Plan sized to a conservative estimate of launch traffic, committing early to secure the lowest published hourly rate before real usage data exists for the application.
Why it is wrong: Tempting because the three-year rate is the deepest, but committing hourly spend against unknown traffic risks paying for unused commitment, and the stem forbids any term commitment until usage is clear.
A regional Standard Reserved Instance, which provides the billing discount and reserves launch capacity in any Availability Zone in the Region.
Why it is wrong: Regional Reserved Instances give the billing discount and zone flexibility but do not reserve capacity in any specific Availability Zone.
Purchase a one-year Compute Savings Plan sized to the nightly peak so the batch fleet is covered by a committed lower hourly rate.
Why it is wrong: A Savings Plan rewards steady baseline usage, but committing to the nightly peak means paying for the commitment during the long idle daytime hours, which wastes money on an intermittent job.

Examworthy is not affiliated with or endorsed by Amazon Web Services. Original, blueprint-aligned practice material only.