Exam Strategy, 30-Day Study Plan, and Glossary
A week-by-week plan to prepare for SAA-C03, a keyword-to-service cheat sheet for scenario questions, exam-day tactics, a glossary of the terms used in this course, and a final review quiz.
A 30-Day Plan
Plan on 1–2 hours a day, with more on weekends for labs. If you already work with AWS daily, compress weeks 1 and 2.
Week by week
Week 1: Format and security. Take a short baseline practice set on day 1 so you know your weak domains. Read the official exam guide and its in-scope service list. Work through Modules 1–2 and do the IAM role and security group labs.
Week 2: Resilience and performance. Modules 3–4. Build the SNS-to-SQS fan-out, create an RDS instance with Multi-AZ and a read replica, and run the Athena Parquet lab. For each service you meet, write one line: what it’s for, how it fails, what it costs.
Week 3: Cost and design. Modules 5–6. Price three designs in the Pricing Calculator. Redesign the ticketing case study for each changed requirement. Skim the Well-Architected Framework pillar summaries.
Week 4: Practice and review. Take at least two full, timed practice exams (65 questions, 130 minutes). After each one, review every question, including the ones you got right, and note why each wrong option is wrong. Spend the remaining days on your weakest domain. Rest the day before.
Good practice sources include the official AWS Skill Builder practice question set and practice exam, and the AWS documentation FAQs for core services.
Keyword Cheat Sheet
Scenario questions use recurring phrases. Each one strongly suggests a service or pattern:
| If the question says… | Think… |
|---|---|
| “decouple,” “buffer,” “absorb spikes” | SQS |
| ”in order,” “exactly once” | SQS FIFO |
| ”send to multiple systems” | SNS + SQS fan-out, or EventBridge |
| ”orchestrate steps with retries” | Step Functions |
| ”real-time stream,” “replay” | Kinesis Data Streams |
| ”load streaming data into S3 with no code” | Amazon Data Firehose |
| ”ad hoc SQL on S3” | Athena (with Parquet and partitions) |
| “shared file system for Linux” | EFS |
| ”SMB,” “Windows,” “Active Directory” | FSx for Windows File Server |
| ”HPC,” “high-throughput file system with S3” | FSx for Lustre |
| ”microsecond reads from DynamoDB” | DAX |
| ”too many database connections from Lambda” | RDS Proxy |
| ”automatic failover in another AZ” | RDS Multi-AZ |
| ”scale reads” | Read replicas, Aurora replicas, caching |
| ”cross-Region DR with RPO of seconds” | Aurora Global Database or DynamoDB global tables |
| ”static IP,” “UDP,” “fast Regional failover” | Global Accelerator |
| ”cache content at the edge” | CloudFront |
| ”block SQL injection,” “rate limit” | AWS WAF |
| ”block an IP at the subnet” | Network ACL |
| ”private access to S3 from a VPC” | Gateway VPC endpoint |
| ”rotate database credentials” | Secrets Manager |
| ”detect threats,” “compromised instance” | GuardDuty |
| ”find PII in S3” | Macie |
| ”restrict what member accounts can do” | SCPs in AWS Organizations |
| ”sign in once to many accounts” | IAM Identity Center |
| ”interruptible,” “fault-tolerant batch,” “cheapest compute” | Spot Instances |
| ”unknown access pattern” (S3) | S3 Intelligent-Tiering |
| ”archive for years, rarely accessed” | S3 Glacier Deep Archive |
Use these as starting points, not reflexes. Always check the answer against every requirement and the qualifier.
Exam-Day Tactics
A pacing plan for the 130 minutes:
- Pace: 130 minutes for 65 questions is 2 minutes each. If a question takes more than 3 minutes, pick your best answer, flag it, and move on.
- Read the last sentence first, then the scenario, then the options (Module 1, Lesson 2).
- Eliminate first. You can usually remove two options quickly; then compare the remaining two on the qualifier.
- Multiple response questions tell you how many to choose. Every choice must be correct.
- Never leave one blank. There’s no penalty for guessing.
- Use the review pass for flagged questions. Change an answer only if you find a specific reason, not a vague feeling.
- Online proctoring: test your system in advance, clear your desk, and plan for check-in time.
Common mistake: memorizing leaked or “brain dump” questions to pass.
It violates the AWS Certification agreement and can get your certification revoked. It also fails you in practice: questions are reworded and refreshed, and an interviewer or customer will ask why. Learn the design reasons, and the questions become recognizable patterns.
Glossary
| Term | Meaning |
|---|---|
| AZ | Availability Zone: one or more isolated data centers within a Region |
| Region | A geographic area containing multiple AZs |
| Multi-AZ | Running redundant resources in more than one AZ for high availability |
| RTO / RPO | Recovery time objective (max downtime) / recovery point objective (max data loss) |
| Pilot light / warm standby | DR strategies with minimal or scaled-down resources running in a recovery Region |
| IAM role | An identity with temporary credentials, assumed by services, users, or other accounts |
| SCP | Service control policy: an Organizations guardrail that limits permissions in member accounts |
| Permissions boundary | The maximum permissions an identity-based policy can grant to a user or role |
| STS | Security Token Service: issues temporary credentials |
| Security group | Stateful, allow-only firewall attached to a network interface |
| Network ACL | Stateless subnet firewall with allow and deny rules |
| VPC endpoint | Private connection from a VPC to an AWS service (gateway or interface type) |
| KMS | Key Management Service: creates and controls encryption keys |
| Loose coupling | Connecting components through queues, topics, or events so each can fail and scale independently |
| Fan-out | One message delivered to many consumers, typically SNS to multiple SQS queues |
| Visibility timeout | How long an SQS message stays hidden after a consumer receives it |
| Dead-letter queue (DLQ) | A queue that collects messages that repeatedly fail processing |
| Read replica | An asynchronous, readable copy of a database used to scale reads |
| Placement group | Controls how EC2 instances are placed on hardware (cluster, spread, partition) |
| Savings Plans | A 1- or 3-year $/hour commitment in exchange for lower compute prices |
| Spot Instance | Spare EC2 capacity at a deep discount that AWS can reclaim with 2 minutes’ notice |
| Lifecycle policy | Rules that move S3 objects to cheaper classes or delete them over time |
| Well-Architected Framework | AWS’s six-pillar framework for reviewing workload design |
| ADR | Architecture decision record: a short note on why a design decision was made |
Key Takeaway
Follow a four-week plan: learn by domain, build small labs, then switch to timed practice exams and fix your weakest areas. Use the keyword cheat sheet as a starting point, but always check every requirement and the qualifier. Learn why designs work. That passes the exam and makes you useful as an architect.
- Take a short baseline practice set today. Score each domain separately.
- Book a tentative exam date about 30 days out. A date on the calendar beats “when I’m ready.”
- Rewrite the 4-week plan above for your schedule, putting extra time on your two lowest-scoring domains.
- Pick five labs from this course to build, and add them to your calendar.
1. A web tier must hand off image processing jobs that take up to 5 minutes, scale workers independently, and not lose jobs if workers fail. What is the BEST design?
SQS buffers jobs durably and decouples the tiers; workers scale on the backlog and messages reappear if a worker fails.
2. A company needs a shared file system for Windows servers that integrates with Active Directory. Which service fits?
FSx for Windows File Server provides SMB file shares with Active Directory integration.
3. Developers in member accounts must never be able to launch resources outside two approved Regions, even if their IAM policies allow it. What should you use?
An SCP denying actions outside approved Regions applies to every principal in the member accounts, regardless of their IAM policies.
4. A steady workload runs 24/7 on the same instance family for the next 3 years. Which purchase option gives the largest discount for it?
A 3-year commitment on a known family earns the deepest discount for steady usage. Spot is cheaper per hour but can be interrupted, which doesn't suit a steady production workload.