Course menuSAA-C03

Exams / SAA-C03 / Study guide / EC2 purchasing options

EC2 purchasing options: SAA-C03 study notes and practice questions

What to know about EC2 purchasing options for the SAA-C03 exam, from the Design Cost-Optimized Architectures domain (20% of the exam guide), then 3 free questions to test it. Part of the SAA-C03 study guide.

Last verified against current AWS documentation ·How we write our questions

What to know for the exam

EC2 pricing questions usually describe a mixed workload. The trick is to split it into parts and pay for each part the cheapest way that still meets its needs.

The options

Option Commitment Discount (up to) Fits
On-Demand None none Unpredictable or short-term use, new workloads
Compute Savings Plans $/hour for 1 or 3 years ~66% Steady use that may change instance family, size, Region or OS, or move to Fargate or Lambda
EC2 Instance Savings Plans $/hour for one instance family in one Region, 1 or 3 years ~72% Steady use that will stay in one family and Region
Standard Reserved Instances Instance type, Region, 1 or 3 years ~72% Same as Instance Savings Plans; can be sold on the RI Marketplace
Convertible Reserved Instances 1 or 3 years, exchangeable ~66% Steady use where you want to exchange instance attributes
Spot Instances None; can be interrupted with a 2-minute warning ~90% Fault-tolerant, flexible work: batch, CI, stateless web tiers, big data
On-Demand Capacity Reservations None (billed whether used or not) none (combine with Savings Plans) Guaranteeing capacity in a specific zone, for example for a known event
Dedicated Hosts On-Demand or reserved varies Per-socket or per-core software licenses, compliance needing a physical server
Dedicated Instances On-Demand or reserved varies Hardware not shared with other customers, without host-level control

Key facts

  • Savings Plans apply automatically to matching usage; Compute Savings Plans also cover Fargate and Lambda.
  • Spot best practices: be flexible across many instance types and zones, use the price-capacity-optimized allocation strategy, and checkpoint work so an interruption only loses a little.
  • Auto Scaling groups can mix On-Demand (for a base) and Spot (for the rest), across several instance types.
  • Capacity Reservations don’t give a discount on their own, but Savings Plans and regional RIs apply to them.

How to choose

  • Steady 24/7 baseline that may move to containers: Compute Savings Plan.
  • Nightly batch jobs that can restart: Spot.
  • Short-term project, a few weeks: On-Demand.
  • License tied to physical cores: Dedicated Hosts.
  • Must have capacity in one zone on a specific date: On-Demand Capacity Reservation.

Common traps

  • Spot for work that can’t be interrupted, such as a single-instance database.
  • Reserved Instances when the workload will change instance family. Compute Savings Plans are more flexible.
  • Dedicated Hosts for “isolation” when Dedicated Instances (or nothing extra) would do. They cost more.

Try the 3 practice questions

A genomics research team runs a batch pipeline every night from 22:00.

It needs about 400 vCPUs for roughly 6 hours and must finish by 07:00, which leaves about 3 hours of slack.

Jobs save a checkpoint every 10 minutes and can resume on any instance. The pipeline runs well on many compute optimized and general purpose instance types, and the capacity is not needed during the day.

Which purchasing approach gives the LOWEST cost while meeting the deadline?

Choose one answer.

  1. A. Use a Spot-only Auto Scaling group spanning many instance types and all Availability Zones, with the price-capacity-optimized strategy.
  2. B. Use a Spot-only Auto Scaling group with one instance type in one Availability Zone, with the lowest-price allocation strategy.
  3. C. Buy a 1-year Compute Savings Plan sized for 400 vCPUs of On-Demand usage, and run the pipeline on On-Demand Instances each night.
  4. D. Run the pipeline on On-Demand Instances that a scheduled scaling action launches at 22:00 nightly and ends when the jobs finish.
Show the answer and explanation

Answer: A

Explanation

Checkpointed, restartable jobs with slack in the schedule are a good fit for Spot Instances. How Spot is requested matters: being flexible across many instance types and Availability Zones, with the price-capacity-optimized strategy, reduces interruptions and capacity shortfalls that could threaten the deadline. Commitments such as Savings Plans suit steady round-the-clock use, not 6 hours a night.

Study

Question 2

Harder · choose 2 ·Link to this question (question 2)

A streaming company runs its customer-facing API on 10 m6i.xlarge instances 24 hours a day and expects this baseline to last at least 3 years.

Within the next year, it plans to move the API to Graviton-based m7g instances, and later possibly to containers on AWS Fargate. A few times a month, unpredictable traffic surges need up to 6 extra instances for a few hours. The API cannot tolerate interruptions.

Which TWO purchasing choices give the LOWEST total cost? (Choose two.)

Choose two answers.

  1. A. Buy a 3-year Compute Savings Plan with an hourly commitment that covers the usage of the 10 baseline instances.
  2. B. Run the surge instances as On-Demand Instances that the Auto Scaling group launches only while they are needed.
  3. C. Buy 3-year Standard Reserved Instances for 10 m6i.xlarge instances to cover the baseline at the deepest discount.
  4. D. Run the extra surge instances as Spot Instances across several instance types to get the largest discount on them.
  5. E. Buy a 3-year EC2 Instance Savings Plan for the m6i family with a commitment that covers all 16 instances.
Show the answer and explanation

Answers: A, B

Explanation

Match each part of the workload to a pricing model. The steady baseline suits a 3-year commitment, and because the company will change instance family and may move to Fargate, the Compute Savings Plan is the commitment that keeps applying. Short, unpredictable surges that cannot be interrupted suit On-Demand, which costs nothing when idle.

Study

A company spends a steady amount every hour on compute and expects this to continue for at least 3 years.

Over that time it plans to move many workloads from Amazon EC2 to AWS Fargate and AWS Lambda, switch EC2 instance families as new generations arrive, and possibly move some workloads to another AWS Region. It wants a discount that follows these changes automatically, without manual exchanges.

Which purchase option should it choose?

Choose one answer.

  1. A. Buy EC2 Instance Savings Plans for the current instance families in the current Region, with a 3-year term.
  2. B. Buy Convertible Reserved Instances with a 3-year term, and exchange them whenever instance families change.
  3. C. Buy Standard Reserved Instances with a 3-year term for the instance types that the company runs today.
  4. D. Buy Compute Savings Plans with a 3-year term and an hourly commitment that matches the steady baseline.
Show the answer and explanation

Answer: D

Explanation

Only Compute Savings Plans apply automatically across instance families and Regions and to Fargate and Lambda, in exchange for an hourly spend commitment over 1 or 3 years. EC2 Instance Savings Plans and Reserved Instances can give bigger discounts, but they are tied to EC2, to a Region, and (except by manual exchange) to an instance family.

Study

Want more on EC2 purchasing options? The 5 full-length practice sets have 10 more EC2 purchasing options questions, 325 in all, none of them in the free exam. One payment of $14.99, no subscription. See the practice sets

Study EC2 purchasing options

More Design Cost-Optimized Architectures topics