AWS SAA-C03 Practice Exam

AWS Certified Solutions Architect Associate SAA-C03

The AWS Certified Solutions Architect - Associate certification validates the ability to design solutions on AWS that are secure, resilient, high-performing, and cost-optimized. The exam covers four domains: Design Secure Architectures, Design Resilient Architectures, Design High-Performing Architectures, and Design Cost-Optimized Architectures. It draws on services such as EC2, S3, VPC, RDS, Aurora, DynamoDB, Lambda, ELB, Auto Scaling, CloudFront, Route 53, IAM, and KMS, and on the AWS Well-Architected Framework as the basis for choosing between them. Candidates are tested on selecting and combining services to meet stated requirements rather than on implementation detail. This certification suits solutions architects and developers with about a year of hands-on AWS design experience.

855 questions in the bank

Question bank updated : questions, answers, explanations and chapters.

Certification Overview

  • Exam name: SAA-C03
  • Vendor: AWS
  • Exam code: SAA-C03
  • Duration: 130 minutes
  • Total questions: about 65
  • Official passing score: 720 on a scale of 100-1,000 (source)
  • Our practice exam pass mark: 72%

Who Should Take This Exam?

Solutions Architect, Cloud Architect, Cloud Engineer, DevOps Engineer, Systems Engineer, Software Developer, Technical Lead, Pre-Sales Solutions Consultant

Prerequisites

No prior certification is required. AWS recommends about one year of hands-on experience designing solutions on AWS.

Topics Covered

  • Design Secure Architectures
  • Design Resilient Architectures
  • Design High-Performing Architectures
  • Design Cost-Optimized Architectures

Question Types

  • Multiple Choice (Single Answer)
  • Multiple Choice (Multiple Answers)

SAA-C03 Practice Questions

Our question bank contains 855+ practice questions for this certification. Sample questions from each exam chapter. Expand a question to see the answer choices. With full access, you get unlimited practice exams with randomized questions from our full question bank.

Design Secure Architectures

A newly created IAM user has no policies attached. What level of access does that user have?
  • A. Read-only access to all AWS services in the account
  • B. Console sign-in only, no programmatic API rights
  • C. Full administrator access to all AWS services
  • D. Zero permissions - IAM enforces deny by default Correct answer

Option D is correct because IAM follows an implicit-deny model: a user with no policies attached has zero effective permissions until an explicit Allow is granted. Options A and C wrongly assume a default access level exists, and B incorrectly narrows the restriction to just programmatic access when in fact console access is blocked too.

Which AWS service inspects HTTP and HTTPS requests and filters them based on rules at the application layer?
  • A. AWS WAF Correct answer
  • B. Amazon GuardDuty
  • C. AWS Shield Advanced
  • D. AWS Network Firewall

Option A is correct because AWS WAF inspects HTTP/HTTPS requests and applies rules (like SQL injection or rate-based rules) at the application layer (layer 7). GuardDuty is a threat detection service, Shield Advanced protects against DDoS at the network/transport layer, and Network Firewall filters traffic more broadly at the VPC level rather than inspecting web application requests.

Which encryption algorithm does S3 use when server-side encryption with S3-managed keys (SSE-S3) is enabled?
  • A. AES-256 Correct answer
  • B. Triple DES
  • C. RSA-2048
  • D. AES-128

Option A is correct because SSE-S3 always encrypts objects using AES-256, with AWS fully managing the keys. Triple DES and RSA-2048 are not used by any S3 server-side encryption option, and AES-128 is not the algorithm SSE-S3 uses.

An IAM policy statement has Effect set to Deny. How does AWS evaluate a request that matches this statement?
  • A. Prompts for MFA before the matched action proceeds
  • B. Blocks the action even if a separate Allow also applies Correct answer
  • C. Logs to CloudTrail but still permits the action
  • D. Forwards the request to root for approval

Option B is correct because in AWS IAM policy evaluation, an explicit Deny always overrides any Allow, regardless of where the Allow comes from. There is no MFA prompt, automatic logging bypass, or root-approval workflow triggered by a Deny statement, which rules out A, C, and D.

A financial services firm needs to ensure all EC2 instances in a VPC can call AWS Systems Manager without internet access. Which TWO resources must be configured? (Choose two)
  • A. A NAT gateway in a public subnet with a 0.0.0.0/0 route from the private subnet
  • B. An S3 gateway endpoint so SSM can download agent updates from S3
  • C. Interface VPC endpoints for ssm, ssmmessages, and ec2messages Correct answer
  • D. Security group rules allowing outbound HTTPS to the endpoint's private IPs Correct answer
  • E. An internet gateway so the SSM agent can reach the public endpoint

Options C and D are correct because reaching Systems Manager without internet access requires interface VPC endpoints for ssm, ssmmessages, and ec2messages, plus security group rules permitting outbound HTTPS from the instances to those endpoints' private IPs. A NAT gateway and an internet gateway would route traffic over the public internet, which the requirement explicitly avoids, and an S3 gateway endpoint isn't required for the core SSM connectivity being described, which rules out A, B, and E.

Design Resilient Architectures

What is the primary benefit of using an Amazon SQS queue between an application tier and a processing tier?
  • A. It automatically encrypts messages in transit
  • B. It synchronizes both tiers to prevent duplicate processing
  • C. It stores messages across multiple Regions for durability
  • D. It lets each tier scale independently by decoupling them Correct answer

Option D is correct because SQS decouples the application and processing tiers, letting each scale independently based on its own load instead of being tied to the other's capacity. It does not synchronize tiers to prevent duplicates, encrypt messages automatically without configuration, or provide cross-Region durability by default, so A, B, and C don't hold up.

An Amazon RDS Multi-AZ deployment automatically fails over to the standby when the primary becomes unavailable. What is the standby instance's role during normal operation?
  • A. The standby serves read traffic to offload the primary
  • B. It receives synchronous replication but serves no traffic Correct answer
  • C. The standby handles scheduled analytics queries
  • D. It replays the primary binlog asynchronously, about 1s behind

Option B is correct because in a Multi-AZ RDS deployment the standby receives synchronous replication from the primary purely for failover readiness and serves no read or write traffic during normal operation. Options A and C are wrong because the standby cannot be queried directly, and option D describes asynchronous replication, which is how read replicas work, not the Multi-AZ standby.

How many Availability Zones does Amazon S3 replicate an object across by default?
  • A. 1 AZ, until cross-region replication is configured
  • B. Every AZ in the Region to reach 11 nines durability
  • C. At least 3 AZs within the same Region Correct answer
  • D. 2 AZs: a primary copy and one backup

Option C is correct because Amazon S3 automatically stores objects redundantly across a minimum of three Availability Zones within a Region to achieve its high durability. Option A is wrong because this replication happens by default with no configuration needed, option B overstates it since S3 does not need every AZ in the Region, and option D describes an outdated two-copy model that is not how S3 achieves durability.

An Application Load Balancer performs health checks on registered EC2 targets. What does the ALB do with a target that fails its health check?
  • A. It terminates the target and launches a replacement
  • B. It keeps routing to the target at reduced connection limit
  • C. It quarantines the target and retries health checks for 24 hours
  • D. It stops routing to that target and uses healthy ones only Correct answer

Option D is correct because an Application Load Balancer simply stops sending new requests to targets that fail health checks and routes only to the healthy targets until the failed one recovers. It does not terminate instances itself, keep routing to unhealthy targets at reduced capacity, or quarantine targets for a fixed period, so A, B, and C are incorrect.

A cloud security vendor needs to design a resilient web application that survives an Availability Zone failure with no manual intervention. Which TWO design choices are required? (Choose two)
  • A. Amazon Inspector detects vulnerabilities that could cause AZ-level failures in advance
  • B. EC2 instances across multiple AZs in an Auto Scaling group, redistributing when one AZ fails Correct answer
  • C. Elastic IPs on each EC2 instance so addresses survive an AZ failure for reconnection
  • D. RDS Multi-AZ so the database fails over automatically to a healthy AZ Correct answer
  • E. Application config on a single EBS volume shared across AZs using EFS mount targets

Options B and D are correct because spreading EC2 instances across multiple AZs in an Auto Scaling group lets the group automatically redistribute capacity when one AZ fails, and RDS Multi-AZ automatically fails the database over to a standby in a healthy AZ, together covering both the compute and data tiers without manual action. Inspector only scans for vulnerabilities and doesn't provide runtime resilience, Elastic IPs don't automatically move to surviving instances on their own, and EBS volumes cannot be shared across AZs via EFS mount targets, so that option describes an invalid configuration.

Design High-Performing Architectures

A bank needs to process real-time clickstream data from millions of users. The data should be ingested at scale and processed by multiple independent consumers. Which AWS service is designed for this pattern?
  • A. Kinesis Data Streams with multiple independent consumer apps Correct answer
  • B. Amazon SQS standard queue with multiple consumers polling in parallel
  • C. Amazon SNS with one SQS queue subscription per consumer
  • D. Amazon MQ with a topic exchange fanning out to bound queues

Option A is correct because Kinesis Data Streams is purpose-built for high-throughput real-time ingestion that multiple independent consumer applications can read from concurrently, each at its own pace. SQS standard queues remove a message once processed, preventing true independent multi-consumer replay, and SNS with per-consumer SQS queues or Amazon MQ are not designed for this scale of continuous streaming analytics.

An application writes data to DynamoDB with high throughput. During predictable business hours the write rate is 10x the off-peak rate. Which DynamoDB capacity mode minimizes cost for this pattern?
  • A. On-demand capacity mode so DynamoDB scales up writes without any pre-provisioning
  • B. Provisioned capacity with DynamoDB Auto Scaling that adjusts write capacity on a CloudWatch target Correct answer
  • C. Provisioned capacity set to the peak rate so writes proceed without throttling throughout the day
  • D. DynamoDB Accelerator (DAX) in front of the table to absorb writes and batch-flush them to DynamoDB

Option B is correct because provisioned capacity with DynamoDB Auto Scaling adjusts write capacity to match the predictable business-hours pattern, avoiding both the per-request premium of on-demand mode and the waste of provisioning for peak all day. On-demand mode is better suited to unpredictable traffic, provisioning fixed at the peak rate pays for unused capacity off-peak, and DAX is a read-side cache that doesn't help with sustained write throughput.

A healthcare provider wants to send application log events from Lambda functions to multiple downstream consumers: an OpenSearch cluster for search, an S3 bucket for archival, and a monitoring dashboard. Which AWS service fans out the log events to all three targets?
  • A. Amazon EventBridge with a single event bus rule routing log events to three targets in the same rule
  • B. Kinesis Data Firehose with three delivery streams each subscribed to the same CloudWatch Logs group
  • C. CloudWatch Logs Insights with three saved queries on a schedule to extract and push data to targets
  • D. Amazon Kinesis Data Streams with one stream and three consumers each forwarding to their destination Correct answer

Option D is correct because streaming the log events into a single Kinesis Data Stream lets multiple independent consumer applications each read the full stream and forward records to their own destination, whether OpenSearch, S3, or a dashboard. EventBridge and Firehose each route matched events or records to their configured targets rather than supporting arbitrary independent consumer applications reading the same feed, and CloudWatch Logs Insights is a query tool, not a fan-out delivery mechanism.

A game studio wants to reduce read latency for its DynamoDB table and handle millions of reads per second. Which TWO techniques directly reduce read latency or increase read throughput? (Choose two)
  • A. Export tables to S3 so Athena absorbs read traffic and reduces load on DynamoDB
  • B. Use eventually consistent reads to consume half the read capacity units per request Correct answer
  • C. Enable DynamoDB Streams so reads are served from the change stream instead of the table
  • D. Create a GSI on the primary key to serve reads from a secondary partition key space
  • E. Add a DAX cluster in front of DynamoDB for in-memory caching of frequently accessed items Correct answer

Options B and E are correct because eventually consistent reads consume half the read capacity of strongly consistent ones, directly easing throughput pressure, and a DAX cluster caches frequently accessed items in memory, serving many reads without hitting the table at all and delivering microsecond latency. Exporting to S3 for Athena doesn't help live read traffic against the table, DynamoDB Streams only captures data changes rather than serving reads, and adding a GSI on the same primary key doesn't create a new query pattern or reduce latency for existing lookups.

A news publisher wants to accelerate global content delivery for a static website hosted in an S3 bucket. Which TWO configurations speed up delivery for users worldwide? (Choose two)
  • A. S3 Requester Pays so users fund transfer costs and company bandwidth is not throttled
  • B. S3 Transfer Acceleration using the CloudFront edge network for faster transfers Correct answer
  • C. S3 Intelligent-Tiering storing frequent-access objects on faster tiers for users
  • D. Cross-Region Replication to 10 regions so users download from the nearest S3 bucket
  • E. CloudFront distribution caching S3 content at edge locations close to users worldwide Correct answer

Options B and E are correct because S3 Transfer Acceleration routes transfers through the CloudFront edge network for faster paths to the bucket, and a CloudFront distribution caches the site's content at edge locations close to users worldwide, directly cutting latency for global visitors. Requester Pays only shifts who pays for transfer costs, Intelligent-Tiering optimizes storage cost rather than delivery speed, and replicating to ten regions would require a custom routing layer to actually direct each user to their nearest bucket, which isn't provided automatically.

Design Cost-Optimized Architectures

A news publisher needs an RDS database that automatically scales storage as data grows without any downtime. Which RDS feature enables this?
  • A. RDS read replicas, which share the primary storage load and increase effective write throughput
  • B. RDS Multi-AZ deployment, which distributes primary storage across Availability Zones for redundancy
  • C. RDS Storage Auto Scaling, which increases allocated storage when free space falls below a threshold Correct answer
  • D. RDS Proxy, which pools connections and reduces storage pressure by limiting concurrent writes

Option C is correct because RDS Storage Auto Scaling automatically increases allocated storage whenever free space drops below a threshold, with no downtime or manual intervention required. Read replicas address read throughput rather than storage capacity, Multi-AZ provides redundancy rather than automatic storage growth, and RDS Proxy manages connection pooling, not storage.

A fintech startup stores objects in S3 across multiple storage classes. They need to ensure objects are automatically moved from S3 Standard to S3 Intelligent-Tiering after 30 days without manual intervention. Which S3 feature does this?
  • A. S3 Storage Lens with automated actions on low access frequency
  • B. S3 Replication rules changing the storage class of replicated objects
  • C. S3 Object Lock governance mode preventing storage class changes after 30 days
  • D. S3 Lifecycle rule transitioning objects to Intelligent-Tiering after 30 days Correct answer

Option D is correct because an S3 Lifecycle rule is the built-in feature for automatically transitioning objects between storage classes, such as from Standard to Intelligent-Tiering, after a set number of days with no manual action needed. Storage Lens only reports on usage patterns rather than acting on objects, replication rules control where copies go rather than storage class transitions over time, and Object Lock enforces retention against deletion or overwrite, not storage class changes.

A hospital network needs to migrate a 10 TB Oracle database to AWS with minimal downtime. Which AWS service enables continuous replication of ongoing changes while the initial full-load migration is in progress?
  • A. AWS DataSync, syncing file system changes to the target RDS instance
  • B. RDS export to S3, capturing a point-in-time snapshot for import
  • C. AWS Snowball Edge, shipping data then re-syncing delta changes on return
  • D. AWS DMS with ongoing replication using CDC alongside the full load Correct answer

Option D is correct because AWS DMS performs an initial full load of existing data and then uses change data capture to continuously replicate ongoing changes made during that migration window, minimizing downtime at cutover. DataSync is built for file and object data rather than transactional databases, exporting a snapshot to S3 only captures a single point in time with no ongoing replication, and Snowball Edge is for offline bulk transfer, not continuous change capture.

A data engineering team needs to migrate data from an on-premises Oracle database to Amazon RDS PostgreSQL with minimal downtime. Which TWO AWS services work together to accomplish this migration? (Choose two)
  • A. AWS DataSync synchronizing Oracle transaction logs to RDS PostgreSQL in real time
  • B. AWS Glue ETL extracting Oracle data, converting to Parquet, and loading into RDS PostgreSQL
  • C. AWS Schema Conversion Tool converting the Oracle schema and stored procedures to PostgreSQL Correct answer
  • D. AWS DMS performing the initial data load and ongoing CDC from Oracle to RDS PostgreSQL Correct answer
  • E. Amazon Athena querying Oracle via JDBC and inserting results into the RDS PostgreSQL target

Options C and D are correct because the AWS Schema Conversion Tool converts the Oracle schema and stored procedures into PostgreSQL-compatible equivalents, and AWS DMS then performs the initial full data load followed by ongoing change data capture from Oracle, keeping the target in sync until cutover with minimal downtime. DataSync isn't built for database transaction log replication, a Glue ETL job converting to Parquet is a batch data lake pattern rather than a live relational migration path, and Athena can't directly query Oracle over JDBC and insert results into RDS the way described.

A ride-sharing platform needs to run a batch job that processes a large dataset and can be interrupted without significant cost, since the job can be retried. Which TWO EC2 purchasing options minimize cost for this interruptible workload? (Choose two)
  • A. EC2 Spot Instances using spare capacity at up to 90% discount, reclaimed with a 2-minute notice Correct answer
  • B. EC2 Reserved Instances with a 1-year no-upfront option providing lower hourly rates than On-Demand
  • C. EC2 Dedicated Hosts for compliance requiring isolated hardware for batch workloads
  • D. EC2 On-Demand in an Auto Scaling group that terminates instances when the batch job completes
  • E. AWS Batch with Spot compute environment, auto-handling interruptions and retrying failed jobs Correct answer

Options A and E are correct because Spot Instances offer savings of up to 90% off On-Demand pricing specifically for workloads that can tolerate interruption with a two-minute warning, and AWS Batch with a Spot compute environment natively handles those interruptions and automatically retries failed jobs, fitting this retry-tolerant batch workload perfectly. Reserved Instances commit to steady, predictable usage rather than being priced around interruption tolerance, Dedicated Hosts exist for compliance and licensing isolation and cost more, not less, and running On-Demand instances in an Auto Scaling group still pays full On-Demand rates regardless of when they're terminated.

Free Study Guide

Read the free AWS SAA-C03 study guide: every exam topic explained in plain English, with diagrams, exam tips and the mistakes to avoid.

More AWS Practice Exams

All AWS practice exams →

Practice with realistic mock exams to prepare for your AWS certification.