intermediateSAA-C038-10 weeks prep8 min read
SAA-C03: AWS Solutions Architect Associate — Study Guide
Complete study guide for the AWS SAA-C03 Solutions Architect Associate exam — the most popular AWS certification. Covers EC2, VPC, S3, RDS, HA architecture, IAM, and cost optimisation.
awssaa-c03solutions-architectintermediateamazonec2vpcs3rdshigh-availability
Domains
9
Key concepts
16
Study time
8-10 weeks
Exam Overview
| Detail | Info |
|---|---|
| Exam code | SAA-C03 |
| Duration | 130 minutes |
| Questions | 65 (scenario-based multiple choice, multiple response) |
| Passing score | 720 / 1000 |
| Cost | $150 USD |
| Validity | 3 years |
| Audience | Cloud architects, engineers designing AWS solutions |
Domain Weightings
| Domain | Weight |
|---|---|
| Design Resilient Architectures | 26% |
| Design High-Performing Architectures | 24% |
| Design Secure Applications and Architectures | 30% |
| Design Cost-Optimised Architectures | 20% |
Domain 1: Resilient Architectures (26%)
High Availability with EC2
- Availability Zones — deploy across ≥2 AZs for HA.
- Auto Scaling Groups (ASG) — maintain desired capacity; replace unhealthy instances; scale in/out based on policies.
- Elastic Load Balancers — distribute traffic across healthy instances.
| ELB Type | Layer | Use case |
|---|---|---|
| ALB (Application) | Layer 7 | HTTP/HTTPS, path/host routing, WebSockets |
| NLB (Network) | Layer 4 | Ultra-low latency, TCP/UDP, static IP |
| GWLB (Gateway) | Layer 3 | Virtual appliances (firewalls, IDS) |
Database resilience
- RDS Multi-AZ — synchronous standby replica; automatic failover in ~1 minute; for HA (not read scaling).
- RDS Read Replicas — asynchronous copies; for read scaling; can be in different regions.
- Aurora — up to 6 copies across 3 AZs, auto-repair; up to 15 read replicas; Global Database for cross-region.
- DynamoDB — serverless NoSQL; multi-AZ by default; Global Tables for active-active multi-region.
Decoupling with messaging
- SQS (Simple Queue Service) — message queue; decouple producers and consumers; at-least-once delivery. Standard (unordered) vs FIFO (ordered, exactly-once).
- SNS (Simple Notification Service) — pub/sub fan-out; sends to multiple subscribers (SQS, Lambda, email, HTTP).
- EventBridge — event bus; route events from AWS services to targets.
Disaster recovery strategies
| Strategy | RTO | RPO | Cost |
|---|---|---|---|
| Backup & Restore | Hours | Hours | Lowest |
| Pilot Light | ~10 min | Minutes | Low |
| Warm Standby | Minutes | Seconds | Medium |
| Multi-Site Active/Active | Near zero | Near zero | Highest |
Domain 2: High-Performing Architectures (24%)
Caching
- ElastiCache (Redis/Memcached) — in-memory cache for database query results, session store.
- CloudFront — CDN; cache static and dynamic content at edge locations globally; TTL-based invalidation.
- DynamoDB Accelerator (DAX) — in-memory cache specifically for DynamoDB; microsecond latency.
Storage performance
| Storage | IOPS | Use case |
|---|---|---|
| EBS gp3 | Up to 16,000 | General purpose, recommended default |
| EBS io2 | Up to 64,000 | High-performance databases |
| EBS st1 | Throughput-optimised | Big data, data warehouses |
| Instance Store | Very high | Ephemeral, temporary data, caching |
| EFS | Scales automatically | Shared file system across EC2 instances |
| FSx for Lustre | Millions of IOPS | HPC, ML training |
Serverless and event-driven
- Lambda — run code without servers; up to 15 minutes; triggers: API Gateway, S3, DynamoDB Streams, SQS, SNS.
- Fargate — serverless containers; no cluster management.
- API Gateway — managed API endpoint; integrate with Lambda, HTTP backends, AWS services.
Domain 3: Secure Architectures (30%)
IAM deep dive
- Policies — identity-based (attached to user/role), resource-based (attached to resource like S3 bucket), session policies.
- Evaluation logic: explicit Deny > explicit Allow > implicit Deny.
- Cross-account access — role in Account B; Account A assumes it via STS AssumeRole.
- Service roles — IAM roles assumed by AWS services (EC2 instance profile, Lambda execution role).
- Permission boundaries — set max permissions; even if policy grants more, boundary caps it.
Network security
- Security Groups — stateful, instance-level, allow-only rules.
- NACLs — stateless, subnet-level, allow + deny rules; rule number evaluated in order.
- VPC Flow Logs — capture IP traffic for audit and troubleshooting.
- AWS WAF — Layer 7 web application firewall (ALB, CloudFront, API Gateway).
- AWS Shield — DDoS protection. Standard (free) + Advanced.
- AWS Network Firewall — stateful, managed firewall for VPC perimeter.
Encryption
- KMS — managed key service; CMKs for fine-grained control.
- S3 encryption — SSE-S3 (AWS manages), SSE-KMS (CMK), SSE-C (customer key), client-side.
- EBS encryption — encrypt at rest with KMS; encrypted snapshots; cannot convert unencrypted to encrypted in-place (copy + encrypt).
- RDS encryption — enable at creation; cannot encrypt existing unencrypted DB (snapshot → copy encrypted → restore).
S3 security
- Block Public Access — account-level guardrail overrides any permissive policy/ACL.
- Bucket policies — resource-based JSON policies; apply to any principal.
- VPC endpoints — S3 gateway endpoint keeps traffic within AWS network.
- MFA Delete — requires root account MFA to permanently delete versioned objects.
Domain 4: Cost-Optimised Architectures (20%)
EC2 pricing models
| Model | Best for | Savings vs On-Demand |
|---|---|---|
| On-Demand | Unpredictable, short-term | Baseline |
| Reserved 1yr | Steady-state workloads | ~40% |
| Reserved 3yr | Very stable long-term | ~60% |
| Savings Plans | Flexible committed spend | ~66% |
| Spot | Fault-tolerant, batch, flexible timing | ~90% |
Storage cost optimisation
- S3 Intelligent-Tiering — auto-tier based on access patterns; no retrieval fee; monitoring fee per object.
- S3 Lifecycle policies — automate tier transitions.
- EBS — use
gp3overgp2(same performance, 20% cheaper); delete unattached volumes. - Snapshots — incremental; use Data Lifecycle Manager for automated rotation.
Architecture cost patterns
- Lambda + API Gateway vs EC2 — Lambda can be cheaper for irregular traffic; EC2 better for constant load.
- NAT Gateway — charged per GB; for very high egress, consider NAT Instance.
- Data transfer — egress to internet charged; within same AZ free; between AZs charged.
- Reserved Capacity — commit to RDS, ElastiCache, Redshift instances for 1–3 years.
Must-Know Services for SAA-C03
Route 53 routing policies
| Policy | Use case |
|---|---|
| Simple | Single resource |
| Weighted | A/B testing, traffic split |
| Latency | Route to lowest-latency region |
| Failover | Primary/secondary with health checks |
| Geolocation | Route by country/continent |
| Geoproximity | Route by distance (with bias) |
| Multi-Value | Up to 8 healthy random records |
S3 essentials
- Versioning — preserve all object versions; MFA delete for extra protection.
- Replication — CRR (cross-region) and SRR (same-region); requires versioning on both buckets.
- Pre-signed URLs — time-limited access without making bucket public.
- Transfer Acceleration — uses CloudFront edge locations to accelerate uploads.
VPC essentials
- Subnets: public (route to IGW) vs private (route to NAT Gateway).
- VPC Peering: connect two VPCs (no transitive routing).
- VPC Endpoints: gateway (S3, DynamoDB — free) and interface (all others — charged).
- Transit Gateway: hub-and-spoke; connect many VPCs and on-premises.
Study Plan (8–10 Weeks)
| Weeks | Focus |
|---|---|
| 1–2 | EC2, IAM, VPC fundamentals |
| 3 | S3, EBS, EFS, storage options |
| 4 | RDS, Aurora, DynamoDB, ElastiCache |
| 5 | Load balancers, Auto Scaling, Route 53 |
| 6 | CloudFront, Lambda, API Gateway, SQS/SNS |
| 7 | Security: KMS, WAF, Shield, IAM policies |
| 8 | Cost optimisation, Savings Plans, Reserved |
| 9–10 | Full practice exams + weak areas |
Key Resources
| Resource | Notes |
|---|---|
| Stephane Maarek (Udemy) | Most popular SAA-C03 course; very comprehensive |
| Adrian Cantrill | Deeper architecture focus; excellent labs |
| Tutorials Dojo | Best practice exams — closest to real exam style |
| AWS Well-Architected Framework | Read all 6 pillars |
| AWS Skill Builder | Free official practice questions |
Common Exam Traps
- Multi-AZ vs Read Replica — Multi-AZ = HA/failover (synchronous). Read Replica = read scaling (asynchronous). Multi-AZ does NOT help with read performance.
- S3 encryption — SSE-S3 = AWS managed, SSE-KMS = you manage the CMK, SSE-C = you provide the key.
- Spot Instance interruption — 2-minute warning. Design for interruptions (checkpointing, SQS queue).
- SQS vs SNS — SQS = one consumer per message (pull). SNS = fan-out to multiple subscribers (push).
- NACL vs Security Group — NACLs are stateless (must allow return traffic explicitly). SGs are stateful.
- RDS encryption — cannot enable encryption on an existing unencrypted RDS instance. Must snapshot → copy (encrypt) → restore.
