AWS Cloud Practitioner Cheat Sheet: Key Facts
Every key takeaway and common trap from the AWS Cloud Practitioner study guide on one page, grouped by chapter. Tap a lesson link to read the full explanation.
77 facts · 10 chaptersFacts checked 2026-11-15
Chapter 1
Cloud Concepts: Benefits & Economics
Key takeaways
- The six advantages of cloud computing each answer a different problem, so match the scenario's pain point to one advantage rather than to the one that sounds best. → The six advantages of cloud computing
- Elasticity grows and shrinks capacity with demand, while scalability only needs to handle growth. → Elasticity, scalability, agility, high availability and global reach
- Pay-as-you-go turns large up-front capital purchases into operating costs paid as resources are used, but it does not guarantee a lower bill. → Cloud economics: fixed versus variable cost and total cost of ownership
- A fair total cost of ownership comparison counts power, cooling, space, staff, hardware refresh and idle capacity on the on-premises side. → Cloud economics: fixed versus variable cost and total cost of ownership
- Bring Your Own License reuses licenses you already own and keeps compliance with you, while License Included bundles the license into the compute price. → Licensing, rightsizing and automation
Common trap
- A popular distractor swaps the definitions, saying scalability releases capacity and elasticity only handles growth. It is the other way round: releasing capacity when demand falls is what makes a design elastic. → Elasticity, scalability, agility, high availability and global reach
- Watch for absolute claims such as cloud always costs less, every bill falls automatically, or the main benefit is a tax break. The accurate answer says savings vary with how resources are used. → Cloud economics: fixed versus variable cost and total cost of ownership
Quick reference
| Advantage | The problem it answers |
|---|---|
| Trade fixed expense for variable expense | Paying a lot up front before you know how much you will use |
| Benefit from massive economies of scale | A small buyer cannot reach low unit costs alone |
| Stop guessing capacity | Forecasts leave servers idle or short at busy times |
| Increase speed and agility | Waiting weeks for hardware makes experiments slow and costly |
| Stop spending money running and maintaining data centers | Staff spend their time racking, powering and replacing servers |
| Go global in minutes | Distant users see slow responses and you have no local facility |
Chapter 2
Cloud Concepts: Well-Architected & Migration
Key takeaways
- The Well-Architected Framework has six pillars: operational excellence, security, reliability, performance efficiency, cost optimization and sustainability. → The six pillars of the Well-Architected Framework
- Pillars can be traded off by business context, but security and operational excellence generally are not. → Pillar trade-offs and the AWS Well-Architected Tool
- AWS CAF has six perspectives (Business, People, Governance, Platform, Security, Operations) and four journey phases (Envision, Align, Launch, Scale). → The AWS Cloud Adoption Framework: outcomes, perspectives and phases
- Rehost moves servers unchanged, replatform adds a managed service without a redesign, and refactor redesigns the application for cloud-native features. → Migration strategies: the 7 Rs
- AWS DMS moves database data, AWS SCT converts schemas between engines, and AWS Application Migration Service moves whole servers. → AWS migration services and tools
Common trap
- Two pairs are easy to confuse. Logging actions to hold people accountable is security (traceability), while collecting telemetry about workload health is operational excellence (observability). Reducing resources to save energy is sustainability, while reducing them to cut spending is cost optimization. → Pillar trade-offs and the AWS Well-Architected Tool
- AWS Elastic Disaster Recovery replicates servers much like AWS Application Migration Service, but its purpose is recovering from an outage, not migrating. Likewise, AWS Storage Gateway gives on-premises apps access to cloud storage; it does not migrate a database. → AWS migration services and tools
Quick reference
| Pillar | Focus and a typical principle |
|---|---|
| Operational excellence | Running workloads well and improving processes: observability, frequent small reversible changes |
| Security | Protecting data and systems: maintain traceability by logging and auditing every action |
| Reliability | Working correctly and recovering from failure: scale horizontally so one failure affects less |
| Performance efficiency | Using resources efficiently to meet requirements: go global in minutes for lower latency |
| Cost optimization | Business value at the lowest price point: attribute spending, stop paying for undifferentiated heavy lifting |
| Sustainability | Lowering environmental impact: maximize utilization so less energy is used per unit of work |
| Strategy | What happens |
|---|---|
| Rehost (lift and shift) | Servers move to AWS unchanged |
| Relocate | A whole platform moves in bulk to the cloud version of the same platform |
| Replatform (lift and reshape) | Some optimization, such as a managed database, with no redesign |
| Refactor (re-architect) | Redesign for cloud-native features; the most complex and costly path |
| Repurchase (drop and shop) | Replace with a different product, often SaaS |
| Retain | Keep it where it is for now, for example because of data residency |
| Retire | Shut it down because it no longer has business value |
Chapter 4
Security: IAM & Access Management
Key takeaways
- Protect the root user with MFA, keep no root access keys, and use root only for the few tasks that require it, such as closing a standalone account. → The root user and sign-in credentials
- IAM user groups hold only IAM users, while roles give people, applications and other accounts temporary credentials instead of long-term keys. → IAM users, user groups and roles
- IAM denies by default, needs an explicit allow, and lets an explicit deny in any policy override every allow. → Policies, deny rules and least privilege
- IAM Identity Center signs in the workforce across many accounts, while Amazon Cognito signs in the customers of web and mobile apps. → Federation, IAM Identity Center, Amazon Cognito and secrets
- AWS Secrets Manager stores and rotates credentials, while Systems Manager Parameter Store holds plain configuration settings that need no rotation. → Federation, IAM Identity Center, Amazon Cognito and secrets
Common trap
- Distractors often put a role or an EC2 instance inside a user group, nest one group in another, or copy IAM users into a second account. None of these is possible. The fix is almost always a role. → IAM users, user groups and roles
- Wrong answers treat read-only access as automatically safe, or let the newest or the user-level policy beat an explicit deny. Safe-sounding broad access is still not least privilege, and an explicit deny always wins. → Policies, deny rules and least privilege
Quick reference
| Identity | What it is | Key limit |
|---|---|---|
| IAM user | One person or workload in one account, with long-term credentials | Belongs to a single account |
| IAM user group | A set of IAM users that share attached policies | Holds only IAM users, not roles or other groups |
| IAM role | A set of permissions that a trusted person, service or account assumes | Gives temporary credentials, no long-term ones |
| Need | Service |
|---|---|
| Employees reach many accounts in an organization with one sign-in from the corporate directory | AWS IAM Identity Center, with permission sets for each job function |
| Customers of a web or mobile app sign up and sign in, including with Google or Apple | Amazon Cognito |
| A managed Microsoft Active Directory on AWS | AWS Directory Service |
| A database password kept out of code and rotated on a schedule | AWS Secrets Manager |
| Plain settings such as endpoints, in a hierarchy, with no rotation | AWS Systems Manager Parameter Store |
Chapter 5
Security: Services & Resources
Key takeaways
- AWS WAF filters web requests by their content, AWS Shield protects against DDoS attacks, and AWS Firewall Manager rolls both out across accounts. → Blocking web attacks and DDoS: AWS WAF, AWS Shield and AWS Firewall Manager
- Every AWS customer gets AWS Shield Standard automatically, while AWS Shield Advanced is a paid subscription for enhanced DDoS protection. → Blocking web attacks and DDoS: AWS WAF, AWS Shield and AWS Firewall Manager
- Amazon GuardDuty detects suspicious activity, Amazon Inspector finds software vulnerabilities, and Amazon Macie discovers sensitive data in Amazon S3. → Detecting threats, flaws and sensitive data: GuardDuty, Inspector and Macie
- AWS Security Hub gathers and prioritizes findings, Amazon Detective investigates their root cause, and AWS Trusted Advisor recommends fixes without making them. → Collecting, investigating and advising: Security Hub, Detective and Trusted Advisor
- Third-party security products are bought in AWS Marketplace, and suspected abuse coming from AWS resources is reported to AWS Trust and Safety. → AWS Marketplace and where to find security help
Common trap
- Answers often swap WAF and Shield, or say Firewall Manager filters traffic. Remember: WAF reads request content, Shield handles floods, and Firewall Manager only deploys and maintains protections such as the other two. → Blocking web attacks and DDoS: AWS WAF, AWS Shield and AWS Firewall Manager
- Distractors mix up the three: GuardDuty for unpatched software, Inspector for a malicious IP address, or Macie for a database table. Match the noun in the question: activity goes to GuardDuty, vulnerability to Inspector, sensitive data in S3 to Macie. → Detecting threats, flaws and sensitive data: GuardDuty, Inspector and Macie
- Do not let similar names fool you. AWS Security Center is a website of resources, while AWS Security Hub is a service that gathers findings inside your accounts. Likewise, AWS Marketplace sells third-party products, while AWS Service Catalog holds a company's own approved products. Only Trust and Safety takes abuse reports. → AWS Marketplace and where to find security help
Quick reference
| Service | What it looks for | Typical finding |
|---|---|---|
| Amazon GuardDuty | Suspicious activity, using threat intelligence and machine learning | An instance contacting known malicious IP addresses or mining cryptocurrency |
| Amazon Inspector | Known software vulnerabilities in EC2 instances, container images in Amazon ECR and Lambda functions | A package with a published vulnerability |
| Amazon Macie | Sensitive data in objects stored in Amazon S3 | Files holding names and account numbers |
| Resource | Use it for |
|---|---|
| AWS Knowledge Center | How-to and troubleshooting articles for common questions |
| AWS Security Blog | In-depth articles and walkthroughs by AWS experts |
| AWS Security Center | The central website for AWS security resources, including how to report a vulnerability |
| AWS Security Bulletins | Announcements of vulnerabilities |
| AWS Trust and Safety | Reporting AWS resources used for abuse, such as port scans or spam, even by non-customers |
Chapter 6
Technology: Deployment & Global Infrastructure
Key takeaways
- The AWS Management Console suits one-off visual tasks, the AWS CLI suits scripts, and AWS SDKs suit application code, and all three call the same AWS service APIs. → Ways to access AWS: console, CLI, SDKs and APIs
- AWS CloudFormation builds repeatable infrastructure from templates, Elastic Beanstalk runs uploaded code, and Systems Manager operates resources that already exist. → Deployment tools and cloud deployment models
- A Region contains multiple isolated Availability Zones, and edge locations are separate sites used for content delivery. → Regions, Availability Zones and high availability
- Deploying across multiple Availability Zones in one Region is the standard way to survive a data center failure, and AWS does not do it for you. → Regions, Availability Zones and high availability
- A second Region is for disaster recovery, data residency and serving distant users, and data is copied between Regions only when you set it up. → Multiple Regions, data residency and edge services
Common trap
- Do not mix deployment models with service models. IaaS and SaaS describe how much a provider manages, not where the application runs, and a VPC is a network inside AWS, not an on-premises data center. → Deployment tools and cloud deployment models
- Wrong answers claim that a bigger instance is highly available, that AWS copies instances between zones on its own, or that only a second Region can survive a data center failure. Multiple Availability Zones are the expected answer. → Regions, Availability Zones and high availability
Quick reference
| Method | Best fit |
|---|---|
| AWS Management Console | A browser interface for one-time or visual tasks, with no code needed |
| AWS CLI | Commands typed in a terminal or saved in shell scripts |
| AWS SDKs | Calling AWS from inside an application, in its programming language |
| AWS CloudFormation | Describing whole environments as code so they can be built again the same way |
Chapter 7
Technology: Compute & Networking
Key takeaways
- Pick the EC2 instance category by the bottleneck: memory optimized for large in-memory data, compute optimized for processor-heavy jobs, storage optimized for heavy local disk reads and writes. → Amazon EC2 instance types, auto scaling and load balancers
- Amazon ECS and Amazon EKS orchestrate containers, Amazon ECR stores images, and AWS Fargate is the serverless compute layer that runs under them. → Containers, serverless compute, Lightsail and AWS Batch
- Auto scaling changes how many instances exist to give elasticity, while a load balancer spreads traffic only to healthy targets. → Amazon EC2 instance types, auto scaling and load balancers
- Security groups are stateful and work at the instance level with allow rules only, while network ACLs are stateless, work at the subnet level and can deny. → Inside a VPC: subnets, gateways, security groups and network ACLs
- AWS Direct Connect is a dedicated private link that is not encrypted by default, while AWS Site-to-Site VPN is an encrypted tunnel over the internet. → Route 53 and connecting networks to AWS
Common trap
- Do not mix up the two jobs. A load balancer never launches instances, and auto scaling never distributes traffic. AWS Compute Optimizer only recommends better sizes; it does not add or remove capacity. → Amazon EC2 instance types, auto scaling and load balancers
- Amazon SQS holds a queue of messages but provisions no compute, so it cannot replace AWS Batch. Likewise, Fargate is not an alternative to ECS; it runs underneath it. → Containers, serverless compute, Lightsail and AWS Batch
- Watch for swapped pairs. Distractors flip stateful and stateless, or instance and subnet, or claim a security group can deny. Another trap says a subnet is public because its security group allows web traffic; the route to an internet gateway is what makes it public. → Inside a VPC: subnets, gateways, security groups and network ACLs
- Private does not mean encrypted. AWS Direct Connect keeps traffic off the internet, but encryption is something the customer adds separately. The two encrypted IPsec tunnels belong to Site-to-Site VPN, which does run over the internet. → Route 53 and connecting networks to AWS
Quick reference
| Category | Clue in the question |
|---|---|
| Memory optimized | Large data sets processed in memory, such as an in-memory database |
| Compute optimized | High-performance processors: batch processing, media transcoding, game servers |
| Storage optimized | Heavy sequential reads and writes on local disks |
| General purpose | An even mix of compute, memory and networking |
| Feature | Security group | Network ACL |
|---|---|---|
| Applies to | An instance | A subnet |
| State | Stateful: return traffic is allowed automatically | Stateless: return traffic needs its own rule |
| Rules | Allow only | Allow and deny |
| Need | Service |
|---|---|
| Consistent performance, traffic off the public internet, from a data center | AWS Direct Connect |
| Encrypted office-to-VPC link set up quickly over the internet | AWS Site-to-Site VPN |
| Individual remote workers on their own laptops | AWS Client VPN |
| One hub for many VPCs and on-premises networks | AWS Transit Gateway |
| Private access to a supported service through a VPC endpoint | AWS PrivateLink |
| Create, publish and manage an API in front of Lambda | Amazon API Gateway |
Chapter 8
Technology: Databases & Storage
Key takeaways
- Amazon RDS takes over OS patching, backups and hardware, but query tuning stays with the customer, and EC2 is the choice only when you need control of the operating system. → Relational databases: Amazon RDS, Aurora and databases on EC2
- A Multi-AZ deployment is for availability with automatic failover, while read replicas are for scaling reads. → Relational databases: Amazon RDS, Aurora and databases on EC2
- When a migration changes the database engine, AWS SCT converts the schema and AWS DMS moves the data. → NoSQL, in-memory and graph databases, plus database migration
- Choose an S3 storage class by how often data is read and how fast it must open, and use Intelligent-Tiering when the pattern is unknown and Lifecycle rules when it is predictable by age. → Amazon S3 object storage, storage classes and lifecycle policies
- EBS volumes survive an instance stop but instance store data does not, and AWS does not back up EBS for you unless you set up snapshots or AWS Backup. → Block and file storage, Storage Gateway, AWS Backup and disaster recovery
Common trap
- In a Multi-AZ DB instance deployment, the standby is not a read replica. Wrong answers say the standby can take read traffic while it waits, or that a backup or snapshot can answer queries. Only read replicas serve reads. → Relational databases: Amazon RDS, Aurora and databases on EC2
- Distractors give AWS DMS the schema job or AWS SCT the data job, or credit planning tools such as Migration Hub and Migration Evaluator with converting or copying data. Keep schema with SCT and data with DMS. → NoSQL, in-memory and graph databases, plus database migration
- Glacier Flexible Retrieval and Deep Archive need a restore before reading, so they are wrong whenever the data must open at once. Intelligent-Tiering never deletes objects; deletion by age is a Lifecycle expiration. → Amazon S3 object storage, storage classes and lifecycle policies
- A common wrong answer claims instance store keeps data through a stop because the data moves to EBS. It does not. Another says Amazon EFS fits Windows SMB shares; Windows, SMB or Active Directory points to FSx for Windows File Server. → Block and file storage, Storage Gateway, AWS Backup and disaster recovery
Quick reference
| Feature | Purpose |
|---|---|
| Multi-AZ DB instance deployment | A standby in another Availability Zone takes over automatically if the primary fails; it does not serve reads |
| Read replicas | Asynchronously updated copies that answer read-only queries and take load off the primary |
| Snapshot | A point-in-time copy that must be restored by hand before use |
| Storage class | Best fit |
|---|---|
| S3 Standard | Frequently read data |
| S3 Standard-IA | Rarely read data that must open at once, kept across several Availability Zones |
| S3 One Zone-IA | Rarely read data you can re-create, kept in one Availability Zone for less |
| S3 Intelligent-Tiering | Unknown or changing access; moves objects between tiers by observed access, with no retrieval fees |
| S3 Glacier Instant Retrieval | Archives read a few times a year that must open immediately |
| S3 Glacier Flexible Retrieval and S3 Glacier Deep Archive | Archives that can wait for a restore; Deep Archive is the lowest-cost storage in AWS, with the slowest restore |
Chapter 9
Technology: AI, Analytics & Other Services
Key takeaways
- Transcribe turns speech into text, Polly turns text into speech, Translate changes the language, Comprehend finds meaning such as sentiment, Textract reads scanned documents and Rekognition analyzes images and video. → AI and machine learning services: pre-trained services and SageMaker AI
- Amazon SageMaker AI is for building, training and deploying your own models, while the pre-trained AI services need no machine learning expertise. → AI and machine learning services: pre-trained services and SageMaker AI
- AWS Glue catalogs and prepares data, Athena queries files in S3 in place, Redshift is the data warehouse, EMR runs Spark and Hadoop clusters, Kinesis collects streams and Amazon Quick builds dashboards. → Analytics services: Glue, Athena, Redshift, EMR, OpenSearch, Kinesis and Amazon Quick
- SQS holds messages until a worker pulls them, SNS pushes one message to many subscribers, EventBridge routes events by rule and Step Functions runs ordered workflows. → Messaging and workflows: SQS, SNS, EventBridge, Step Functions, Amazon Connect and SES
- WorkSpaces gives a full cloud desktop, AppStream 2.0 streams single applications and WorkSpaces Secure Browser gives only a remote browser. → Developer tools, cloud desktops, AWS Amplify and AWS IoT Core
Common trap
- Pair questions punish reversed directions. Polly and Transcribe are opposites, and Lex holds the conversation while Polly only speaks. Translate never measures opinion, and Textract reads documents, not audio. → AI and machine learning services: pre-trained services and SageMaker AI
- Occasional queries on files in S3 point to Athena; the same reports every week on years of structured data point to Redshift. Amazon SQS is a work queue, not a streaming pipeline, so it is wrong for clickstreams. → Analytics services: Glue, Athena, Redshift, EMR, OpenSearch, Kinesis and Amazon Quick
- Pull versus push is the key. SNS does not hold messages for one worker, and SQS does not broadcast. SNS can send a simple email notification, but a full email service on your own domain is SES. AWS CodePipeline has stages too, but it releases software, not business processes like Step Functions. → Messaging and workflows: SQS, SNS, EventBridge, Step Functions, Amazon Connect and SES
Quick reference
| Service | Input to output |
|---|---|
| Amazon Transcribe | Speech to text |
| Amazon Polly | Text to lifelike speech |
| Amazon Translate | Text in one language to another language |
| Amazon Comprehend | Text to meaning, such as positive or negative sentiment |
| Amazon Textract | Scanned documents to text, forms and tables |
| Amazon Rekognition | Images and video to labels, including unsafe content flags |
| Amazon Lex | Conversations: chatbots and voice bots that understand intent |
| Need | Service |
|---|---|
| A full, persistent Windows desktop joined to Active Directory | Amazon WorkSpaces |
| One desktop application streamed to a browser, no data on the device | Amazon AppStream 2.0 (now called Amazon WorkSpaces Applications in AWS documentation) |
| Internal websites and SaaS apps only, no VPN, no software on the laptop | Amazon WorkSpaces Secure Browser |
| Full-stack web and mobile app with Git-based deploys and pull request previews | AWS Amplify |
| Connect thousands of devices securely and route their messages | AWS IoT Core |
Chapter 10
Billing, Pricing & Support
Key takeaways
- On-Demand fits short or unpredictable work that must not be interrupted, Spot fits flexible work that can be stopped, and Savings Plans or Reserved Instances fit steady use you can commit to. → EC2 purchasing options: On-Demand, Spot, Savings Plans, Reserved Instances and dedicated hardware
- Capacity Reservations and zonal Reserved Instances reserve capacity in one Availability Zone, while regional Reserved Instances and Savings Plans give a billing discount without reserving capacity. → EC2 purchasing options: On-Demand, Spot, Savings Plans, Reserved Instances and dedicated hardware
- Pricing Calculator estimates before you build, Cost Explorer analyzes past spend and forecasts, Budgets alerts on thresholds and Cost and Usage Reports give the most detailed line items. → Cost management tools: Pricing Calculator, Cost Explorer, Budgets, Cost and Usage Reports and Compute Optimizer
- Consolidated billing gives one bill, combines usage for volume pricing and shares Reserved Instance and Savings Plans discounts across accounts by default. → Data transfer costs, consolidated billing and cost allocation tags
- Business Support+ is the recommended minimum for production, Enterprise Support adds a designated TAM and Unified Operations adds incident management engineers for mission-critical workloads. → AWS Support plans, Trusted Advisor, AWS Health and where to get help
Common trap
- Capacity and discount are different things. Distractors claim a Capacity Reservation gives a discount, that a regional Reserved Instance reserves capacity, or that Reserved Instances renew themselves. At expiry, an instance simply keeps running at On-Demand rates. → EC2 purchasing options: On-Demand, Spot, Savings Plans, Reserved Instances and dedicated hardware
- Pricing Calculator never reads your live account, so it cannot chart past bills or alert on spending. Cost Explorer analyzes but does not send threshold alerts, and Amazon Quick can visualize billing data but does not produce it. → Cost management tools: Pricing Calculator, Cost Explorer, Budgets, Cost and Usage Reports and Compute Optimizer
Quick reference
| Option | When it fits |
|---|---|
| On-Demand Instances | No commitment; short, spiky or unpredictable work that must not be interrupted |
| Spot Instances | Lowest cost for flexible, interruption-tolerant work; AWS can reclaim the capacity |
| Compute Savings Plans | Hourly spend commitment that applies across instance families, Regions, EC2, Fargate and Lambda |
| EC2 Instance Savings Plans | Commitment locked to one instance family in one Region |
| Standard or Convertible Reserved Instances | 1-year or 3-year term; Convertible can be exchanged for other attributes for a smaller discount, Standard cannot be exchanged |
| On-Demand Capacity Reservations | Guaranteed capacity in one Availability Zone with no fixed term; not a discount |
| Dedicated Hosts | A whole physical server with socket and core visibility for bring-your-own per-core licenses |
| Tool | What it answers |
|---|---|
| AWS Pricing Calculator | What will a planned architecture cost, upfront, monthly and yearly, before we build it |
| AWS Cost Explorer | What did we actually spend over past months, what is the trend and what is the forecast |
| AWS Budgets | Alert us by email or Amazon SNS when actual or forecasted cost or usage crosses a threshold |
| AWS Cost and Usage Reports | The most detailed line items by resource, hour and tag, delivered to our own S3 bucket for Athena or Amazon Quick |
| AWS Compute Optimizer | Which resources are oversized, based on actual utilization metrics |