SAA-C03 Questions 51-60: Cost-Optimized Architectures

Designing cost-optimized architectures is worth 20% of SAA-C03. These questions rarely ask for the cheapest option at any price; they ask for the lowest cost that still meets the stated requirements for performance, access time, and availability. Storage classes, purchase options, network paths, and capacity modes are the main levers.

This is the sixth set in our SAA-C03 series (Questions 51-60). It follows SAA-C03 Questions 41-50 on high-performing architectures. Purchase options such as Reserved Instances and Spot are covered in Questions 1-10. Every question is an original practice scenario mapped to the official exam guide, not a copy of a real exam item.

For each scenario, note the access pattern and the requirement that cannot be broken, then pick the cheapest option that still meets it.

Question 51: Committing to Spend While Moving to Containers

A company spends a steady amount per hour on compute. Over the next year it plans to move some workloads from EC2 to AWS Fargate and AWS Lambda, and to change instance families. It wants a discount for committing to that spend without locking into one instance family or Region. What should it buy?

A) Standard Reserved Instances for its current instance types

B) EC2 Instance Savings Plans for one instance family

C) A Compute Savings Plan

D) Spot Instances for all workloads

👁 Reveal Answer

Correct Answer: C

Explanation: Savings Plans give lower prices in exchange for a commitment to a consistent amount of usage, measured in dollars per hour, for one or three years. Compute Savings Plans are the most flexible: they apply to EC2 usage regardless of instance family, size, operating system, or Region, and also to Fargate and Lambda. A and B lock the discount to specific instance types or a family in a Region, which does not suit the planned changes. D can be interrupted and is unsuitable for steady workloads that cannot tolerate interruption.

Reference: Savings Plans types

Question 52: Logs That Age Out of Use

An application writes logs to S3. Logs are read often for 30 days, rarely for the rest of the first year, almost never after that, and must be deleted after seven years. Retrieval of logs older than a year can take up to 48 hours. How should storage be managed at the lowest cost?

A) Keep everything in S3 Standard and delete it by hand after seven years

B) An S3 Lifecycle configuration that moves logs to S3 Standard-IA after 30 days, to S3 Glacier Deep Archive after one year, and expires them after seven years

C) Store all logs in S3 Glacier Deep Archive from the first day

D) Store logs on EBS volumes and take snapshots

👁 Reveal Answer

Correct Answer: B

Explanation: S3 Lifecycle rules can transition objects to cheaper storage classes as they age and expire them when they are no longer needed. Standard-IA suits data that is accessed less often but needs fast access, and Glacier Deep Archive is the lowest-cost class for data that can wait hours for retrieval. Automatic expiration removes logs at seven years. A pays Standard prices for old data and depends on someone remembering to delete. C makes recent logs, which are read often, slow and expensive to retrieve. D is more expensive and harder to manage for this kind of data.

Reference: Transitioning objects using Amazon S3 Lifecycle

Question 53: Nobody Knows How Often the Data Is Used

A research platform stores millions of datasets in S3. Some are used heavily for weeks and then ignored, and others sit untouched for months before suddenly becoming popular. Access patterns cannot be predicted. Which storage class minimizes cost without risking retrieval fees or slow access?

A) S3 Intelligent-Tiering

B) S3 One Zone-IA

C) S3 Glacier Flexible Retrieval

D) S3 Standard only

👁 Reveal Answer

Correct Answer: A

Explanation: S3 Intelligent-Tiering automatically moves objects between access tiers based on how they are actually used, and moves them back to the frequent access tier when they are accessed again. It charges a small monthly monitoring and automation fee per object but no retrieval charges, which makes it the recommended class for unknown or changing access patterns. B is cheaper for infrequent data but has retrieval charges and stores data in only one Availability Zone. C requires a restore before objects can be read. D is simple but overpays for data that goes cold.

Reference: Managing storage costs with Amazon S3 Intelligent-Tiering

Question 54: Rarely Opened, but Needed Immediately

A hospital keeps medical images that are opened about once a quarter, usually during a patient’s follow-up visit. When a doctor opens one, it must load in milliseconds. The images must be kept for years. Which storage class is the most cost-effective?

A) S3 Glacier Deep Archive

B) S3 Glacier Flexible Retrieval

C) S3 Standard

D) S3 Glacier Instant Retrieval

👁 Reveal Answer

Correct Answer: D

Explanation: S3 Glacier Instant Retrieval is designed for long-lived data that is rarely accessed, such as once a quarter, but needs millisecond retrieval. It offers much lower storage cost than S3 Standard, with higher per-request retrieval costs and a minimum storage duration, which suits this pattern. A and B require restoring objects first, which takes minutes to hours, so doctors could not open images immediately. C meets the speed requirement but costs more for data that is rarely used.

Reference: Understanding S3 Glacier storage classes

Question 55: A Huge NAT Gateway Bill

EC2 instances in private subnets read and write several terabytes per day to S3 in the same Region. Cost Explorer shows that most of the bill is NAT gateway data processing charges. What change reduces the cost the most?

A) Add a gateway VPC endpoint for S3 and update the private subnets’ route tables

B) Replace the NAT gateway with a larger one

C) Move the instances to a public subnet with public IP addresses

D) Turn on S3 Transfer Acceleration

👁 Reveal Answer

Correct Answer: A

Explanation: NAT gateways charge per hour and per gigabyte of data processed. A gateway endpoint for S3 routes S3 traffic directly from the private subnets to S3 over the AWS network, so it no longer passes through the NAT gateway, and gateway endpoints have no additional charge. B does not exist as a cost fix; NAT gateways are not sized that way and still charge per gigabyte. C weakens security. D adds charges and is meant for long-distance uploads.

Reference: Gateway endpoints (Amazon VPC)

Question 56: Cutting Costs in a Test Environment

A company copies its production VPC design, with one NAT gateway in each of three Availability Zones, into a development environment used only by internal testers. Developers can tolerate a short loss of outbound internet access if one Availability Zone fails. How can the company reduce the development environment’s cost?

A) Remove all NAT gateways and give every instance a public IP address

B) Keep three NAT gateways but move them to private subnets

C) Use a single NAT gateway that all private subnets in the development VPC route through

D) Add a fourth NAT gateway for redundancy

👁 Reveal Answer

Correct Answer: C

Explanation: A NAT gateway per Availability Zone gives the best resilience but costs more, because each gateway has an hourly charge. When a workload can tolerate losing outbound access during an Availability Zone failure, as in this development environment, a single shared NAT gateway is a reasonable trade-off. Instances in other zones also pay for cross-zone data transfer, which is usually small for test traffic. A exposes instances to the internet. B breaks NAT, which must sit in a public subnet. D increases cost.

Reference: NAT gateway use cases

Question 57: A Server That Waits All Day

A finance team runs a script that takes about three minutes, a dozen times a day, whenever a file arrives. Today it runs on an EC2 instance that is on 24 hours a day and is idle most of the time. What is the most cost-effective option?

A) A larger EC2 instance so the script finishes faster

B) An AWS Lambda function triggered when the file arrives

C) A Reserved Instance for the existing server

D) A dedicated host

👁 Reveal Answer

Correct Answer: B

Explanation: Lambda charges only for the number of requests and the compute time used while the function runs, so a short job that runs a few times a day costs a small fraction of an always-on server. Triggering it from the file arrival (for example, an S3 event notification) also removes the need for polling. The three-minute run time is well within Lambda’s 15-minute maximum. A still pays for idle time. C lowers the hourly rate but still pays for every idle hour. D is the most expensive option.

Reference: AWS SAA-C03 exam guide, Task 4.2 (cost-optimized compute)

Question 58: Which Instances Are Too Big?

A company has 300 EC2 instances that teams sized years ago. Finance suspects many are over-provisioned. The company wants recommendations for better instance types and sizes based on actual CPU, memory, and network use, without building its own analysis. Which service fits?

A) AWS CloudTrail

B) Amazon Inspector

C) AWS Shield

D) AWS Compute Optimizer

👁 Reveal Answer

Correct Answer: D

Explanation: AWS Compute Optimizer analyzes utilization metrics and classifies resources such as EC2 instances as under-provisioned, over-provisioned, or optimized, with recommended instance types and estimated savings. Memory recommendations need memory metrics from the CloudWatch agent. A records API activity. B finds software vulnerabilities. C protects against DDoS attacks.

Reference: Get EC2 instance recommendations from Compute Optimizer

Question 59: Steady Traffic on an On-Demand Table

A DynamoDB table was created in on-demand capacity mode when the application launched. A year later, its traffic is high, steady, and predictable throughout the day, and the bill has grown. What change is most likely to reduce cost?

A) Add DynamoDB Accelerator (DAX) to every request

B) Switch the table to provisioned capacity mode with auto scaling, sized for the known traffic

C) Add more global secondary indexes

D) Enable DynamoDB Streams

👁 Reveal Answer

Correct Answer: B

Explanation: On-demand mode is ideal for new, unpredictable, or spiky workloads because you pay per request and plan no capacity. When traffic becomes steady and predictable, provisioned capacity with auto scaling usually costs less, and reserved capacity can lower it further. A adds a cache cluster cost and mainly helps read-heavy repeated access. C adds write and storage costs. D adds stream read costs and does not reduce the table’s capacity charges.

Reference: Evaluate your DynamoDB table’s capacity mode

Question 60: Paying Too Much to Send Data to Users

A software company lets users download installers from an S3 bucket, and data transfer out to the internet has become its largest cost. Downloads are the same files requested many times. Which change reduces data transfer cost while also improving download speed?

A) Serve the files through Amazon CloudFront with the bucket as the origin

B) Move the bucket to S3 Glacier Flexible Retrieval

C) Enable S3 Cross-Region Replication to a second Region

D) Copy the files to EBS volumes on a fleet of EC2 instances

👁 Reveal Answer

Correct Answer: A

Explanation: With CloudFront, data transferred from AWS origins such as S3 to CloudFront edge locations is not charged, popular files are cached and served from the edge, and CloudFront’s data transfer pricing to viewers is generally lower than serving the same traffic directly from S3. Users also download from a location near them. B makes downloads slow and adds retrieval charges. C doubles storage and adds replication transfer costs. D adds compute and storage costs and still pays internet data transfer.

Reference: Amazon CloudFront pricing

Study Tips for Cost-Optimized Architectures

• Match the storage class to the access pattern. Unknown patterns mean Intelligent-Tiering. Rare access with millisecond retrieval means Glacier Instant Retrieval. Archives that can wait hours mean Glacier Flexible Retrieval or Deep Archive.
• Automate aging. Lifecycle rules transition and expire objects without manual work.
• Know the commitment options. Compute Savings Plans are the most flexible, and Reserved Instances and EC2 Instance Savings Plans give deeper discounts with less flexibility.
• Network paths cost money. Gateway endpoints remove NAT processing charges for S3 and DynamoDB, and CloudFront reduces internet egress costs.
• Pay for what you use. Lambda for sporadic work, Compute Optimizer for right-sizing, and the right DynamoDB capacity mode for your traffic.

Keep Practicing for SAA-C03

You have now covered all four SAA-C03 domains. Review Questions 11-20 for security or Questions 41-50 for performance. The official SAA-C03 exam guide lists every task in scope.