SAA-C03 Questions 41-50: High-Performing Architectures

Designing high-performing architectures is worth 24% of SAA-C03. These questions ask you to pick the right caching layer, database feature, storage type, network service, or event-driven pattern for a performance problem. The trap is usually an option that would help a little but does not address the actual bottleneck.

This is the fifth set in our SAA-C03 series (Questions 41-50). It follows SAA-C03 Questions 31-40 on resilient architectures. Every question is an original practice scenario mapped to the official exam guide, not a copy of a real exam item.

Before choosing, identify where the slowness actually is: the network distance, the database, the storage, or the connection handling.

Question 41: Slow Image Loads for Users Around the World

A news site stores images and videos in an S3 bucket in us-east-1. Readers in Asia and Europe complain that media loads slowly, while US readers are fine. What is the best way to improve performance for global readers?

A) Serve the content through an Amazon CloudFront distribution with the bucket as its origin

B) Move the bucket to a larger storage class

C) Turn on S3 Versioning

D) Add more EC2 instances in us-east-1

👁 Reveal Answer

Correct Answer: A

Explanation: CloudFront is a content delivery network that caches content at edge locations close to viewers, so readers in Asia and Europe are served from nearby locations instead of crossing the world to us-east-1 every time. Using origin access control also lets you keep the bucket private. B is not a performance setting; storage classes affect cost and retrieval. C keeps object versions. D adds compute but does nothing about the distance between readers and the content.

Reference: AWS SAA-C03 exam guide, Task 3.4 (edge networking services such as CloudFront)

Question 42: Large Uploads From Far-Away Users

Film crews around the world upload large video files, often several gigabytes each, directly to a single S3 bucket in us-west-2. Uploads from distant countries are slow and sometimes time out. What should the architect enable?

A) S3 Intelligent-Tiering

B) A larger EBS volume on the crews’ laptops

C) S3 Transfer Acceleration on the bucket, with uploads sent to the accelerated endpoint (using multipart upload for large files)

D) Cross-Region Replication

👁 Reveal Answer

Correct Answer: C

Explanation: S3 Transfer Acceleration routes uploads through the nearest edge location and then across the optimized AWS network to the bucket, which speeds up long-distance transfers of large objects. Combining it with multipart upload makes large files faster and more reliable, because parts upload in parallel and failed parts can be retried. A changes storage cost. B has nothing to do with the network path. D copies objects after they arrive, which does not speed up the upload.

Reference: Getting started with Amazon S3 Transfer Acceleration

Question 43: The Same Queries, Over and Over

A product catalog runs on Amazon RDS for MySQL. Monitoring shows high CPU on the database, and most queries are identical reads for popular product pages that change only a few times a day. What should the architect add to cut database load and response times?

A) A larger RDS instance with more storage

B) An Amazon ElastiCache cluster that caches query results, with the application checking the cache first

C) A Multi-AZ standby

D) AWS Global Accelerator

👁 Reveal Answer

Correct Answer: B

Explanation: Amazon ElastiCache provides an in-memory cache. With a lazy-loading pattern, the application looks in the cache first and only queries the database on a miss, then stores the result with a time to live. Repeated reads of rarely changing data are served in microseconds to milliseconds, and database CPU drops. A treats the symptom at a higher cost. C improves availability, but the standby does not serve reads. D speeds up network paths, not database queries.

Reference: Amazon ElastiCache (Choosing an AWS NoSQL database)

Question 44: Reports Slowing Down the Shop

An online shop’s RDS for PostgreSQL database slows down every afternoon when the analytics team runs heavy read-only reports against it. The reports can use data that is a few seconds old. What is the best fix?

A) Enable Multi-AZ and point the reports at the standby

B) Run the reports at night only

C) Store the reports in S3 Glacier

D) Create a read replica and point the reporting tools at the replica’s endpoint

👁 Reveal Answer

Correct Answer: D

Explanation: RDS read replicas use asynchronous replication to keep a read-only copy of the database, and each replica has its own endpoint. Sending read-heavy reporting there removes that load from the primary, and a few seconds of replication lag is acceptable for these reports. A does not work, because a Multi-AZ DB instance standby cannot serve read traffic. B delays the business need instead of solving it. C stores files; it does not run queries.

Reference: Working with DB instance read replicas

Question 45: A Global Multiplayer Game Over UDP

A game company runs game servers behind Network Load Balancers in three Regions. Players worldwide connect over UDP. The company wants lower and more consistent latency, two fixed IP addresses that players’ firewalls can allow, and fast failover to another Region if one becomes unhealthy. Which service fits?

A) Amazon CloudFront

B) Route 53 simple routing

C) AWS Global Accelerator

D) S3 Transfer Acceleration

👁 Reveal Answer

Correct Answer: C

Explanation: AWS Global Accelerator provides static anycast IP addresses that route TCP and UDP traffic onto the AWS global network at the nearest edge location, then to healthy endpoints such as Network Load Balancers in multiple Regions, with fast failover based on health checks. A is optimized for HTTP and HTTPS content delivery and caching, not UDP game traffic. B does not provide static IPs or traffic acceleration and depends on DNS caching for failover. D is for uploads to S3.

Reference: What is AWS Global Accelerator?

Question 46: A Database That Needs Very High IOPS

A self-managed database on EC2 needs sustained, very high IOPS with consistent sub-millisecond latency for its data volume, well beyond what a general purpose gp3 volume provides. Which EBS volume type should the architect choose?

A) Provisioned IOPS SSD (io2 Block Express)

B) Throughput Optimized HDD (st1)

C) Cold HDD (sc1)

D) General Purpose SSD (gp2) with a smaller volume

👁 Reveal Answer

Correct Answer: A

Explanation: Provisioned IOPS SSD volumes are designed for I/O-intensive, latency-sensitive workloads such as large databases, and io2 Block Express offers the highest performance and durability of the EBS volume types. gp3 includes a 3,000 IOPS baseline and can be provisioned higher, but io2 is the choice for the most demanding sustained workloads. B and C are HDD types built for large sequential throughput or infrequent access, not random IOPS. D would lower performance, because gp2 IOPS scale with volume size.

Reference: Amazon EBS volume types

Question 47: Microsecond Reads From DynamoDB

A gaming leaderboard stored in Amazon DynamoDB receives millions of identical read requests per minute during tournaments. DynamoDB already responds in single-digit milliseconds, but the team needs microsecond read latency without rewriting its data access code much. What should it add?

A) DynamoDB Streams

B) A Global Secondary Index

C) An S3 bucket in front of the table

D) DynamoDB Accelerator (DAX)

👁 Reveal Answer

Correct Answer: D

Explanation: DAX is a fully managed, in-memory cache built for DynamoDB. It is API-compatible with DynamoDB, so applications switch to the DAX client with few code changes, and cached reads return in microseconds instead of milliseconds. It also reduces read load on the table during spikes. A captures item changes for downstream processing. B supports different query patterns, not faster repeated reads. C is not a cache for DynamoDB.

Reference: In-memory acceleration with DynamoDB Accelerator (DAX)

Question 48: Lambda Opens Too Many Database Connections

An API built on AWS Lambda writes to an Amazon RDS for MySQL database. During traffic bursts, thousands of concurrent function instances each open their own connection, the database runs out of connections, and requests fail. What should the architect add?

A) A larger Lambda memory setting

B) Amazon RDS Proxy between the functions and the database

C) A second Lambda function that retries failed requests

D) An SQS FIFO queue in front of the database

👁 Reveal Answer

Correct Answer: B

Explanation: RDS Proxy maintains a pool of database connections and shares them among many application connections, so bursts of Lambda invocations no longer exhaust the database’s connection limit. It also improves failover handling and can use IAM authentication and Secrets Manager for credentials. A increases each function’s resources but not the number of database connections. C adds even more connections. D could buffer writes but requires redesigning the API and adds latency, while RDS Proxy fixes the connection problem directly.

Reference: Amazon RDS Proxy

Question 49: Click Data From Millions of Users in Real Time

A streaming service collects click events from millions of devices. Several applications must process the same events within seconds: one for real-time recommendations, one for fraud detection, and one that archives to S3. The data should be retained for a period so applications can replay it. Which service fits best?

A) Amazon SQS standard queue

B) Amazon RDS

C) Amazon S3 with daily batch jobs

D) Amazon Kinesis Data Streams

👁 Reveal Answer

Correct Answer: D

Explanation: Kinesis Data Streams captures and stores large volumes of streaming data in real time. Multiple consumer applications can read the same stream independently, records are ordered within a shard, and data is retained for a configurable period so consumers can replay it. A delivers each message to one consumer, which then deletes it, so several applications cannot each process every event. B is not built for this ingest rate. C introduces a delay of a day.

Reference: What is Amazon Kinesis Data Streams?

Question 50: Thumbnails Without Servers

Users upload photos to an S3 bucket. Each upload should trigger the creation of a thumbnail within seconds. Uploads arrive unpredictably, from zero to thousands per minute, and the team does not want to manage servers. What is the best design?

A) A cron job on an EC2 instance that scans the bucket every hour

B) An EC2 Auto Scaling group that polls the bucket continuously

C) An S3 event notification that invokes an AWS Lambda function for each new object

D) Amazon Athena queries run by hand

👁 Reveal Answer

Correct Answer: C

Explanation: S3 event notifications can invoke Lambda directly (or route through Amazon EventBridge or SQS) when objects are created. Lambda scales automatically with the number of uploads, runs only when there is work, and requires no servers to manage, which makes this a classic event-driven, serverless pattern. A introduces up to an hour of delay. B keeps servers running even when nothing is uploaded. D analyzes data with SQL; it does not process images.

Reference: AWS SAA-C03 exam guide, Task 3.2 (serverless and event-driven compute)

Study Tips for High-Performing Architectures

• Find the bottleneck first. Distance means CloudFront or Global Accelerator. Repeated reads mean caching. Heavy reporting means read replicas.
• Edge services differ. CloudFront caches HTTP content, Global Accelerator speeds TCP and UDP with static IPs, and S3 Transfer Acceleration speeds uploads to S3.
• Caches by database. ElastiCache in front of relational databases, DAX in front of DynamoDB.
• Connections are a limit too. RDS Proxy solves connection storms from Lambda.
• Streams vs queues. Kinesis lets many consumers read and replay the same data; SQS gives each message to one consumer.

Keep Practicing for SAA-C03

Go back to SAA-C03 Questions 31-40 for resilience, then finish with Questions 51-60 on cost-optimized architectures. The official SAA-C03 exam guide lists every task in scope.