AWS Certified Cloud Practitioner (CLF-C02) Course
New to AWS? Learn cloud concepts, core services, security and billing for CLF-C02 through short lessons and exercises with feedback on every answer.
Inside this course
Exercises you will solve
Services with the most lessons
- Amazon EC2
- Amazon S3
- AWS Identity and Access Management (IAM)
- Amazon Athena
- Amazon Lex
- Amazon Transcribe
- AWS IAM Identity Center (AWS Single Sign-On)
Try a lesson
Restore the database replication service
Fill the gap with the service that performs the full load and captures ongoing database changes.
Amazon RDSHow a lesson works
- ExplanationSeven approaches for different needs
- Sort the statementsSeparate a move from a targeted improvement
- ExplanationCopy the database while it stays in use
- Trace a requestFollow a captured change
- ExplanationReplication reduces downtime, not responsibility
- Complete the diagramRestore the database replication serviceThe exercise you just tried
- ExplanationConnect the migration choice to the data path
- Build the diagramRebuild the replatforming data path
- ExplanationWhat to remember
- 5 to 10 short steps, one at a time.
- Feedback at once: an exercise says what was wrong as soon as you answer.
- No videos: you read and solve at your own pace.
Course scope
The course follows the official CLF-C02 exam guide and covers all 4 domains and 19 tasks.
Define the benefits of the AWS Cloud2 free21 lessons
Pay-as-you-go pricing means paying for cloud resources according to measured billing units instead of buying infrastructure in advance. Units can include virtual machine runtime, storage kept over time, requests processed, and data transferred. Each service defines what it measures.
Total cost of ownership (TCO) is the full cost of providing and operating a workload over a chosen period. Owned infrastructure costs more than its purchase price: facilities, electricity, cooling, hardware replacement, software licenses, and staff time all contribute.
Capacity is the amount of work your computing resources can handle. Demand often changes: an online store may be busy during a promotion and quiet afterward. With fixed capacity, buying enough servers for the expected peak leaves spare resources idle during quieter periods. Buying too little can leave the application unable to keep up.
Delivery agility means shortening the time between an idea and a working deployment. A team can try a change, learn whether it works, and use that feedback to improve the next version. Less time spent waiting for infrastructure means more time available for learning and delivery.
An AWS Region is a separate geographic area where AWS provides infrastructure. You can deploy an application in a suitable Region without building and operating your own data center there. This makes entering a new market easier than first establishing local facilities.
High availability means keeping an application usable despite component failures. Moving an application to AWS gives you access to infrastructure for building this resilience, but migration alone does not create it. One virtual machine can still be a single point of failure: if it stops, the application stops.
AWS global infrastructure provides places around the world to run applications and deliver content. Two of its building blocks are Regions and Availability Zones. They describe different levels of the infrastructure, not interchangeable names for a location.
A Region is a geographic area where you deploy AWS resources. For user-facing applications, placing the workload near its users can reduce latency, the time requests take to travel and receive a response. Geographic distance is a useful starting point, but measure response times rather than assuming the nearest Region is always fastest.
Amazon CloudFront is a content delivery network (CDN): a service that delivers content through a worldwide network of locations. A viewer is someone requesting content, such as an image, a video, or a web page.
AWS Global Accelerator improves the performance and availability of applications used around the world. This lesson follows a standard accelerator, which routes traffic to application endpoints based on factors including health and network performance.
Amazon CloudFront is a content delivery network, or CDN, for HTTP and HTTPS content. Choose it when your main goal is delivering websites, images, downloads, videos, or APIs to users around the world.
High availability is the objective of keeping a service usable for the required proportion of time, including when failures occur. It describes the outcome users need: being able to sign in, read information, or complete a purchase. It does not mean that interruptions are impossible.
High availability means keeping an application usable, including when infrastructure fails. A running server is not enough: users must still be able to complete the actions they need. Redundancy provides additional components that can do the same job if one fails.
Elastic Load Balancing (ELB) distributes incoming application traffic across registered targets: resources that handle the traffic. Instead of choosing a server, users connect to the load balancer, which selects a target for each request.
High availability means keeping a service usable for the required proportion of time, including when failures occur. The emphasis is continuity: users should still be able to do useful work despite a disruption. Availability is judged from the user's perspective, not just by whether a server is running.
An AWS Region is a separate geographic area containing multiple Availability Zones. Redundancy within one Region protects against some local failures, but a disruption affecting that entire Region can still make the application unavailable. Deploying in separate Regions gives the application another location from which to serve users.
Elasticity means adjusting resource capacity as demand changes, in both directions. Capacity is the amount of work your resources can handle. An elastic application gains capacity when more work arrives and releases capacity when that extra capacity is no longer needed.
Horizontal scaling changes the number of resources serving a workload. Adding instances or workers is scaling out; removing them is scaling in. This is a scaling mechanism: it describes how capacity changes. The term alone does not tell you what triggered the change.
Amazon EC2 Auto Scaling adjusts the number of EC2 instances running an application. An Auto Scaling group treats those instances as one collection to manage. Adding and removing instances as demand changes provides elasticity: more capacity when needed and less unused capacity afterward.
Elasticity means matching resource capacity to changing demand, in both directions. When more work arrives, capacity increases; when demand falls, unneeded capacity is released. This helps support performance without keeping all peak capacity running continuously.
AWS CloudFormation provisions and manages AWS resources from a text file called a template. You write the template in JSON or YAML to describe the resources you want and their settings, rather than manually creating each resource. This approach is called infrastructure as code.
Identify design principles of the AWS Cloud6 lessons
The AWS Well-Architected Framework provides a repeatable way to evaluate a cloud workload against technical and business requirements. A workload includes the application, infrastructure, data, operating procedures, and people delivering a business capability. Reviewing resource settings alone would miss questions such as who responds to failures.
The AWS Well-Architected Framework provides questions, best practices, and design principles for evaluating a cloud workload. A workload includes the application, infrastructure, data, dependencies, and operating procedures that deliver a business capability. The review asks whether these support the intended business outcomes—not whether the architecture deserves a passing grade.
A workload is a set of components that together deliver business value, such as an online store and its supporting systems. A workload review asks how that workload is designed and operated, and where its architecture could improve.
Resource efficiency means turning infrastructure resources into useful outcomes without unnecessary consumption. Useful work might be a completed order, a successful request, or a finished report. An application that completes the same work with fewer compute hours is more resource-efficient, provided it still meets its requirements.
Right-sizing means selecting resources whose capacity fits the work an application needs to do. Capacity includes processing power, memory, storage performance, and network capacity—not just the size of a virtual machine.
AWS Compute Optimizer uses machine learning to analyze resource configuration and historical utilization: measurements of how resources have been used over time. It recommends configurations that better match the workload. Supported resources include Amazon EC2 instances, Amazon EBS volumes, and AWS Lambda functions. For example, it can recommend a different EC2 instance type based on the instance’s observed demand.
Understand the benefits of and strategies for migration to the AWS Cloud20 lessons
The AWS Cloud Adoption Framework (AWS CAF) helps an organization prepare for cloud adoption as a business transformation, not just a change of infrastructure. A capability is an organizational ability, such as developing workforce skills or responding to incidents. CAF is guidance for assessing and improving these abilities—not a deployable AWS service or a guarantee of results.
A cloud migration moves applications, data, and operational responsibilities—not just servers. AWS organizes this work into three phases: assess, mobilize, and migrate and modernize. They provide a planning structure, not a rule that every application must progress at the same time.
Migration phases organize the work of moving workloads to the cloud. A workload includes an application, its data, and the resources it needs to run. Migration work includes preparation, security, staff readiness, testing, and operational ownership—not just copying servers.
AWS Application Discovery Service has been in maintenance since October 7, 2025. It is closed to new customers; existing customers can keep using it. AWS's migration guide is at https://docs.aws.amazon.com/application-discovery/latest/userguide/application-discovery-service-availability-change.html.
AWS Migration Evaluator helps build a migration business case: an evidence-based comparison of current on-premises costs with estimated costs on AWS. It supports the decision about whether to migrate; it does not move workloads or promise a particular saving.
AWS Application Discovery Service has been in maintenance since 2025-10-07: it is closed to new customers, while existing customers can keep using it. This lesson explains its established role, not a recommendation for new workloads. AWS's migration guide: https://docs.aws.amazon.com/application-discovery/latest/userguide/application-discovery-service-availability-change.html.
AWS Migration Hub has been in maintenance since November 7, 2025. It is closed to new customers; existing customers can keep using it. This lesson describes its established tracking role, not a recommendation for new workloads. AWS's migration guide is at https://docs.aws.amazon.com/migrationhub/latest/ug/migrationhub-availability-change.html.
AWS Prescriptive Guidance is an official AWS technical resource containing recommended approaches drawn from cloud project experience. It helps with migration, modernization, and other cloud projects. It is not a managed service: reading a publication does not provision infrastructure or move a workload.
The AWS Partner Network (APN) is an ecosystem of organizations that provide software and services for customers using AWS. It is not a cloud runtime service. Customers turn to partners when they need specialist products, implementation skills, or extra delivery capacity.
AWS Professional Services provides consultants who work with customer teams toward agreed cloud project outcomes. It is assistance delivered by people through a consulting engagement, not a cloud resource provisioned through an API.
The AWS Partner Network (APN) is an ecosystem of independent organizations. When you engage an AWS Partner for migration assistance, you engage a separate company, not an AWS consulting team. APN itself is not a team that performs your migration.
The AWS Cloud Adoption Framework (AWS CAF) helps an organization prepare for cloud adoption. It organizes readiness into six perspectives: Business, People, Governance, Platform, Security, and Operations. Each perspective highlights capabilities the organization needs—not just technology it needs to buy.
The seven Rs describe choices for a workload—an application and the resources it needs. Rehost moves it with minimal changes, useful when moving quickly matters. Replatform makes targeted improvements without a major redesign, such as moving a self-managed database to a managed database service. Refactor or rearchitect substantially changes the application design to meet needs such as greater flexibility or scale.
The 7 Rs of Migration are choices about what to do with a workload: an application or component that performs useful work. They are not seven phases to complete in order. Choose a strategy for each workload; different components of one application can follow different paths.
Database replication maintains a copy of database data. During a migration, the existing database is the source, and the database being prepared is the target. The source can remain active while data moves.
AWS Application Migration Service, often called MGN, helps move applications from existing servers to Amazon EC2. It supports rehosting: moving the server's operating system, applications, and data with minimal application changes, rather than redesigning the application.
AWS Database Migration Service (AWS DMS) moves database data from a source to a target. The source can keep serving applications during much of the migration, reducing the time applications must stop using it.
A server contains more than database data: it has an operating system, installed software, configuration, and files. AWS Application Migration Service, often called MGN, helps move that server environment to Amazon EC2. This is rehosting: moving an existing workload to AWS with minimal changes rather than redesigning it.
AWS Schema Conversion Tool (AWS SCT) is a tool you install to help prepare a database for a different database engine. Engines can use different SQL syntax, data types, and features. Moving to another engine can therefore require translating database definitions, not simply copying them.
A database contains both data and definitions that organize and use it. Customer records are data. Table definitions, views, and database code such as stored procedures describe how that data is stored or handled.
Understand concepts of cloud economics23 lessons
A fixed cost stays broadly unchanged over a particular period and capacity range, even when usage changes. An annual data-center lease is an example: serving fewer requests does not reduce the agreed rent. A variable cost changes with consumption, such as a charge based on the amount of data stored.
Rightsizing means matching a resource’s type, capacity, and configuration to measured workload needs. The goal is to remove paid-for capacity that adds no useful value while still meeting service requirements, such as response time and availability.
A cheaper storage option saves money only if it still meets the application's needs. Start with the storage form: object, block, or file. These provide different ways to access data, so they are not interchangeable just because one has a lower advertised price.
Total Cost of Ownership (TCO) is the complete cost of a technology solution over an agreed period. It helps answer whether running a workload in the cloud costs less overall than running it on premises—not just whether one server has a lower price.
AWS Pricing Calculator produces a planning estimate from the services you choose and the usage you expect. Start by selecting a service, then enter its Region, configuration, and usage assumptions. The calculator applies supported pricing dimensions to those inputs and shows estimated charges.
Capital expenditure (CapEx) is spending on assets expected to provide benefits over multiple accounting periods. In infrastructure, examples include buying servers, storage equipment, and networking equipment. These are long-lived assets, purchased to support work over time.
Operating expenditure, or OpEx, includes the ongoing costs of running cloud workloads. Those expenses can contain both fixed costs and variable costs. Paying a cloud bill every month does not mean every charge changes with consumption.
Capital expenditure (CapEx) is spending to acquire or improve a long-lived asset. A server purchase that qualifies for capitalization is recorded as an asset rather than charged entirely as an expense immediately. Its cost is generally recognized as expense over its useful life through depreciation.
Bring Your Own License (BYOL) means using software licenses acquired separately from the cloud service. An entitlement is the right to use software under an agreement. Reusing existing entitlements can avoid buying additional software rights, but you still pay for the cloud resources you consume.
License-included pricing means that a selected cloud offering supplies the right to use specified licensed software as part of its price. You pay for the covered software rights through that offering rather than supplying a separate license for that product.
A software license gives permission to use software under stated terms. Paying for a cloud server does not, by itself, give you permission to run every commercial software product on it.
An entitlement is a right to use software under specified terms. Moving software to the cloud does not automatically move that right. A successful installation proves that the software runs, not that your existing licenses cover it. The vendor’s license agreement determines which products, editions, versions, and cloud environments are eligible.
A software license gives you rights to use software under an agreement. License portability is the contractual permission to use an existing license when a workload moves between environments, such as from an on-premises data center to AWS. Being able to copy or install the software does not establish that permission.
A software license gives you permission to use software under a vendor’s terms. With bring your own license (BYOL), you use licenses your organization already holds for an eligible AWS deployment. You must check that the vendor agreement allows that use, including any restrictions on where the software runs.
Resource utilization describes how much available capacity a workload uses. Rightsizing means matching the type, size, and quantity of resources to the workload's needs. It can mean choosing something smaller, adding capacity, or choosing a different balance of resources—not simply choosing the cheapest option.
Rightsizing means matching provisioned resources to measured workload needs with minimal waste. The right configuration must still meet required performance and availability. It may mean reducing capacity, adding capacity, choosing a different resource type, or changing the number of resources.
Determinism means that identical inputs, initial state, and execution conditions produce identical results. For example, an invoice calculation can return the same total each time when the amounts, tax rate, rounding rules, and relevant conditions stay the same.
Infrastructure as code (IaC) means describing infrastructure in reusable files and using a deployment tool to create and update it. Instead of repeating console clicks to build each environment, you keep a definition of the resources and their settings.
Capacity is the amount of work an application’s resources can handle. Demand changes: a busy period needs more capacity, while a quiet period needs less. Keeping enough capacity for the busiest period all the time can leave paid resources idle.
A volume discount is a lower price per unit tied to the quantity of eligible resources consumed or purchased. Cloud providers can share efficiencies from operating at large scale through lower prices. However, benefiting from those efficiencies is not the same as qualifying for a particular discount.
Usage-based pricing means charges are tied to actual, measured consumption rather than simply to having access to a service. The service measures a billable quantity, such as requests processed, compute running time, or storage used over time. This answers the question: How much did you consume?
AWS Organizations brings multiple AWS accounts into an organization. With consolidated billing, the management account pays the combined charges for its own usage and the usage of the member accounts.
Consolidated billing in AWS Organizations brings the charges for an organization's accounts together for payment. The management account is responsible for paying those charges. Member accounts can still run their own workloads; they do not have to move resources into one account.
Understand the AWS shared responsibility model1 free9 lessons
The AWS shared responsibility model divides security between AWS and its customers. AWS handles security of the cloud: protecting the infrastructure that runs AWS services, including physical facilities, hardware, underlying networking, and the virtualization layer.
Amazon EC2 provides virtual servers called instances. You can choose an operating system and install software, much as you would on a server you own. AWS supplies the underlying computing infrastructure, so you do not have to buy or maintain physical servers.
The shared responsibility model assigns duties according to which layers AWS operates and which layers you operate. A useful question is: who manages the software that needs this patch? The answer depends on the service, not just on whether the workload runs on AWS.
Amazon Relational Database Service (Amazon RDS) helps you set up, operate, and scale relational databases. A relational database organizes data into tables of rows and columns. Relationships connect those tables: an order, for example, can refer to a customer through a customer ID.
Patch management means keeping software updated with fixes for security vulnerabilities and reliability problems. In the cloud, patching is a shared responsibility: AWS and customers patch different software layers, rather than both patching the same components.
Configuration management means defining how a system should be configured and keeping it aligned with that approved state. The approved settings form a baseline. They can cover software versions, network access rules, permissions, and database settings. Comparing actual settings with the baseline helps reveal unexpected changes, called configuration drift.
A technology stack includes facilities, hardware, virtualization, operating systems, runtimes, applications, and data. A runtime is the software environment that executes application code. Cloud service models describe which layers the provider operates and which the customer manages.
A database engine is the software that stores data and runs database queries. You can use a supported relational database engine through Amazon RDS, a managed database service, or install and operate database software on Amazon EC2, a virtual server service.
AWS Lambda runs your code without requiring you to provision or manage servers. You deploy a function, configure what invokes it, and Lambda supplies the environment in which it runs. Serverless means the servers are managed for you, not that there are no servers.
Understand AWS Cloud security, governance, and compliance concepts36 lessons
Compliance means meeting applicable requirements and demonstrating that they are met. Those requirements can come from laws, industry standards, contracts, or internal policies. Evaluating compliance asks: Does what we actually do meet what is required?
Requirement applicability means deciding which obligations apply to your organization, its data, and its activities. Requirements can come from laws, industry standards, contracts, and internal policies. Moving a workload to AWS does not decide which requirements apply.
Compliance means meeting applicable requirements and being able to demonstrate that the required controls work. A control is a safeguard or practice, such as restricting access or reviewing permissions. Using AWS provides infrastructure protections, but it does not automatically make a customer application compliant.
An organizational requirement describes an outcome the organization needs. A control is a measure used to achieve or check that outcome. For example, “protect customer records” gives a direction, but it does not tell an engineer exactly what to configure.
AWS Config records the configurations of supported AWS resources so you can understand their settings and how those settings change. You choose which resource types to record. Recording provides the facts that configuration checks and later audits can use.
AWS CloudTrail and AWS Config both provide evidence for auditing, but they describe different things. CloudTrail records activity: what action happened, who performed it, and when. Config records resource configuration: what settings a resource had and how those settings changed.
AWS Control Tower helps establish and govern a multi-account AWS environment. Separate accounts give teams and workloads their own boundaries, but configuring each account by hand can produce inconsistent security settings. Control Tower automates common setup so accounts start from an agreed foundation.
AWS Organizations centrally manages multiple AWS accounts. A management account manages the organization, and other accounts join as member accounts. You can group accounts into organizational units (OUs), such as development and production, to manage accounts with shared policy needs together.
Security protects information and the systems that use it. Confidentiality means only authorized parties can access information. Integrity means information and configurations are protected against unauthorized changes. Availability means systems remain accessible when needed.
Encryption converts readable data, called plaintext, into an unreadable form called ciphertext. Its main purpose is confidentiality: keeping the contents from being understood by someone who should not read them.
Encryption turns readable data into unreadable data called ciphertext. Recovering the readable data requires the appropriate cryptographic key. Where the data is determines which protection you need.
Amazon S3 stores data as objects. An object contains the data itself, such as an image, a backup file, or a log file, together with metadata that describes it. Applications can store and retrieve these objects without managing storage servers.
Amazon CloudWatch Logs collects and stores records from applications, operating systems, and supported AWS services. A log records something that happened, such as a failed sign-in or a rejected network connection. Keeping these records centrally helps you investigate without opening each source machine.
AWS CloudTrail records AWS account activity as events. It helps answer who performed an action, what they did, when, and from where. Actions can originate from the AWS Management Console, command line tools, SDKs, or AWS services.
Amazon CloudWatch helps you understand how resources and applications are running. It collects operational data such as metrics and logs. Use it to investigate rising CPU usage, application errors, or slow responses, and to monitor conditions that need attention.
AWS Artifact provides access to AWS security and compliance documents. You use these documents to understand how AWS has been assessed against particular standards and to supply evidence about AWS during your organization's compliance reviews.
Data residency describes where data resides geographically. A permitted boundary might be a country, an economic area, or an approved set of AWS Regions. The boundary comes from requirements—not simply from whichever location an engineer prefers.
Amazon GuardDuty is a managed threat detection service. Once enabled, it continuously analyzes supported AWS activity for signs of malicious or unauthorized behavior, helping you notice potential threats without manually reviewing every activity record.
Amazon Inspector is an automated vulnerability assessment service. It helps you find security weaknesses in supported workloads so you can address them before they are exploited. A software vulnerability is a known flaw in software, such as an installed package or application dependency.
Amazon GuardDuty detects signs of threats by analyzing activity in your AWS environment. It helps answer, “Is something suspicious happening?” For example, communication with a known malicious destination can indicate that a resource needs investigation.
Amazon Macie and Amazon Inspector identify different kinds of security concerns. Choose between them by asking what you need to discover: sensitive data or workload weaknesses.
An AWS environment can have several kinds of security issues: suspicious activity, vulnerable software, and resource settings that do not meet security best practices. Reviewing each source separately makes it harder to understand the overall picture.
Security posture means the overall state of your security: how resources are configured and where weaknesses need attention. Use AWS Security Hub for a consolidated security view that helps you review and prioritize issues. Its security findings aggregation and configuration checks are available through Security Hub CSPM, which stands for cloud security posture management.
A distributed denial of service (DDoS) attack uses traffic from many sources to overwhelm an application or its network connections. Legitimate users may see slow responses or be unable to connect. The main target is availability: whether people can use the application when they need it.
AWS WAF is a web application firewall. It inspects HTTP and HTTPS requests and helps filter unwanted traffic before your application handles it. It can examine request details such as a URL path or an HTTP header, rather than just a network connection.
A distributed denial of service (DDoS) attack tries to make an application unavailable by overwhelming it with traffic from many sources. AWS Shield provides managed DDoS protection: its central purpose is to help keep applications available during these attacks.
AWS WAF filters web requests reaching supported AWS resources, such as an Amazon CloudFront distribution. Its rules inspect HTTP requests and determine how to handle them, including allowing or blocking them. A web ACL is the collection of rules and settings used for that protection.
Encryption transforms readable data, called plaintext, into unreadable ciphertext using a cryptographic key. Decryption recovers the readable data with the necessary key. The distinction between encryption in transit and encryption at rest is when and where that protection applies.
Encryption in transit protects data as it travels between endpoints. An endpoint is where a protected connection starts or ends. Secure protocols help prevent people along the path from reading the data or changing it without detection.
Encryption at rest turns stored, readable data into unreadable ciphertext. A cryptographic key is needed to recover the readable data. The protection is useful because obtaining the stored bytes alone should not reveal their contents.
Key management protects the secret values that make encryption work. It covers creating keys, storing them safely, controlling their use, and eventually retiring them. Strong encryption offers little protection if an unauthorized person can use its keys.
Encryption in transit protects data while it travels between systems. TLS, the security protocol behind HTTPS, protects a connection against eavesdropping and undetected changes. This is different from encryption at rest, which protects stored data.
AWS Certificate Manager (ACM) manages TLS certificates. A TLS certificate binds an identity, such as a website’s domain name, to a public key. It helps a client check which server it is connecting to when establishing an HTTPS connection.
Encryption turns readable data, called plaintext, into unreadable ciphertext. A cryptographic key controls that operation. Protecting the key matters because someone who obtains it may be able to decrypt the data.
A hardware security module (HSM) is specialized hardware that protects cryptographic keys and performs operations such as encryption and signing. Both AWS CloudHSM and AWS Key Management Service (AWS KMS) use HSMs. The difference is not hardware protection versus no hardware protection.
Amazon CloudWatch monitors resources and applications. A metric is a series of numerical measurements over time. For example, an EC2 instance's CPUUtilization metric records how much of its CPU capacity is being used.
Identify AWS access management capabilities26 lessons
AWS Identity and Access Management (IAM) controls access to AWS resources. An IAM user is an identity in an AWS account. Creating a user does not, by itself, give that user permission to use AWS resources.
The AWS account root user is the identity associated with the email address used to create an AWS account. It is not an IAM user. For an account with root sign-in credentials, you sign in as root using the account's root email address and password, with MFA for additional protection.
Authentication means proving which identity is making a request. An IAM user can have a password for browser sign-in, access keys for programmatic requests, both, or neither. Successful authentication does not itself grant permission to use a resource.
An IAM role is an identity with permissions that trusted identities can use temporarily. Unlike an IAM user, a role has no long-term password or access keys of its own. Assuming a role creates a role session with credentials that expire.
AWS Identity and Access Management (IAM) manages IAM identities, including users and roles, and the policies that define their permissions. A permission describes an action that is allowed or denied on AWS resources. IAM is where you define access, such as allowing an application to read particular files in Amazon S3.
AWS IAM Identity Center provides centrally managed access for your workforce: employees and other people working for your organization. Choose it for routine human access to AWS accounts and business applications, especially when people need access across multiple accounts. Its organization instance supports AWS account access across an AWS organization.
Authentication verifies who is making a request. AWS authenticates both people and workloads, such as applications running on servers. Each presents credentials: evidence that AWS can validate. Use individual identities rather than shared sign-ins so activity can be tied to its source.
Access control governs who may do what with a resource. It includes establishing a requester's identity and enforcing rules about what that requester may do. A requester can be a person or an application, not just someone signing in through a browser.
Privileged Access Management (PAM) means controlling powerful access so that its use is intentional and traceable. The AWS account root user is a particularly sensitive identity because it can perform certain account-level tasks that other identities cannot.
Least privilege means giving an identity the permissions needed for its work, without unrelated access. It limits what mistakes or a compromised account can affect. Start by describing the work, rather than choosing a policy that sounds convenient.
AWS IAM Identity Center, formerly AWS Single Sign-On, helps employees and other workforce users reach assigned AWS accounts and applications. Its AWS access portal gives them a central place to start, rather than a separate sign-in for every destination.
AWS Directory Service helps provide or connect to Microsoft Active Directory, often called AD. A directory holds identities such as users and groups and supports directory features that compatible computers and applications need. AWS Managed Microsoft AD is a managed Active Directory offered through Directory Service.
Amazon Cognito helps developers add sign-up and sign-in to web and mobile applications. Choose it when people need identities for an application you build, such as a shopping app or a customer booking website. These people are application users: they use your product rather than manage its AWS infrastructure.
A credential helps prove who is making a request. An IAM user can have a console password for signing in and, separately, access keys for programmatic requests from tools or applications. An access key consists of an access key ID and a secret access key. Changing the console password does not replace or disable those keys.
An IAM account password policy sets password rules for IAM users in one AWS account. An IAM user with a password can use it to sign in to the AWS Management Console. Account-wide rules give those users a shared baseline instead of leaving password requirements to each person.
A secret is sensitive information such as a database password, an API key, or an access token. Hardcoding means putting that value directly into application code. Anyone who can read a copied source file could then obtain the credential.
Parameter Store, a capability of AWS Systems Manager, stores configuration as named values called parameters. An application can retrieve a database password by its parameter name instead of keeping the password in its source code.
Credential rotation replaces an authentication secret and retires the old one. It coordinates three places: the validating system that accepts the credential, the secret store holding its value, and the applications that use it. A replacement must work before a planned rotation is considered successful, and the old credential must stop working.
AWS Secrets Manager and AWS Systems Manager Parameter Store can both store sensitive values securely. A password being secret does not, by itself, mean that only Secrets Manager can store it.
Multi-factor authentication (MFA) requires evidence from at least two different factor categories: something you know, such as a password; something you have, such as an authenticator device; or something you are, such as a biometric. A stolen password alone is then not enough to complete authentication.
Identity federation lets AWS accept evidence of authentication from an external identity provider, such as a company's corporate sign-in system. Employees can use their existing identities instead of having a separate IAM user in every AWS account. Their corporate passwords do not need to be copied into AWS.
AWS Security Token Service (AWS STS) lets an authorized caller obtain temporary credentials for an IAM role. Assuming a role creates a role session: the caller uses that session's access instead of borrowing another person's long-term keys.
Temporary credentials let a person or application make AWS requests for a limited time. Their permissions determine what the session can do; their expiration determines how long its credentials work. Unlike long-term IAM user access keys, they expire automatically.
The AWS account root user is the account identity associated with its root email address. It has highly privileged access, including account capabilities that ordinary administrators do not have. It is not an IAM user, and attaching an IAM permissions policy is not a way to limit its permissions.
A directory holds information about an organization's users, groups, and computers. Microsoft Active Directory (AD) also provides authentication: checking that someone signing in is who they claim to be. A workforce identity represents someone such as an employee or contractor.
Amazon Cognito helps web and mobile applications manage their users. Authentication establishes who a user is; authorization determines what that user may do. Signing in does not, by itself, grant access to AWS resources.
Identify components and resources for security8 lessons
Amazon VPC provides a logically isolated virtual network for AWS resources. Security groups and network access control lists (network ACLs) filter network traffic. Their rules describe permitted traffic using details such as protocol, port, and source or destination address.
AWS Firewall Manager lets a security team define security policies centrally instead of managing each account separately. It uses AWS Organizations, which groups AWS accounts and organizes them into organizational units, or OUs.
AWS Marketplace is a digital catalog for finding, buying, deploying, and managing third-party software, data, and services. Customers can search by need, compare listings, and review the seller, delivery method, pricing, and terms. Security offerings include products such as third-party firewalls and vulnerability management tools.
AWS Security Center is a public entry point to AWS security information and resources. It helps you discover guidance about protecting cloud workloads, understanding security responsibilities, and finding more detailed material. Think of it as a guide to information, not a view into your AWS account.
AWS Knowledge Center is an official collection of articles answering recurring technical questions and explaining how to troubleshoot specific problems. It is a public information resource, not a service you deploy. You can read its public articles without purchasing a paid AWS Support plan.
The AWS Security Blog is an official AWS publication with security explanations and practical walkthroughs. It covers topics such as identity, data protection, threat detection, and incident response. It is a learning resource, not a deployed service: reading an article does not configure or protect your account.
AWS Trusted Advisor evaluates AWS resources against best-practice checks. A check looks for a particular condition, such as overly open network access or an underused resource, and reports opportunities to improve the environment. It helps you find issues without manually inspecting every resource yourself.
Amazon Macie is a managed service that discovers sensitive data in Amazon S3 objects. An object is a stored file and its metadata. Macie helps you find information such as personal identifiers and financial data so you can decide how to protect it.
Define methods of deploying and operating in the AWS Cloud1 free17 lessons
The AWS Management Console is a browser-based interface for AWS. It lets you explore services, inspect resources, and submit changes through pages and forms. It is useful for learning a service or carrying out an interactive task without writing commands or code.
The AWS Command Line Interface (AWS CLI) is a tool you run in a terminal to operate AWS services. A command names a service, an operation, and any parameters. For example, aws s3api list-objects-v2 --bucket YOUR_BUCKET asks Amazon S3 to list objects in a bucket.
The AWS Management Console is a browser-based interface with pages, menus, forms, and resource lists. The AWS Command Line Interface (AWS CLI) lets you request AWS operations by typing commands in a terminal. They are different ways to manage AWS resources, not separate sets of resources.
AWS Elastic Beanstalk deploys applications and manages the AWS resources that run them. You supply application code and configuration choices instead of setting up every infrastructure resource separately.
AWS CloudFormation and AWS Elastic Beanstalk both automate work in AWS, but they start with different things. CloudFormation starts with a description of the infrastructure you want. Elastic Beanstalk starts with an application you want to run.
An application programming interface (API) is a contract for interacting with software. An AWS service exposes named operations: actions such as reading an object or describing a resource. Each operation defines the parameters it accepts, the successful response it can return, and possible errors. You can use the operation without knowing how AWS implements it internally.
An AWS Software Development Kit (SDK) is a library that application code uses to call AWS services. Choose an SDK for your programming language: for example, Boto3 for Python or the AWS SDK for JavaScript. Your application calls library methods rather than building AWS API requests by hand.
An AWS SDK is a library that application code uses to call AWS services. SDKs support languages including Python, Java, JavaScript, and .NET languages. Boto3 is the AWS SDK for Python.
An application programming interface (API) is a contract between software components. A service's API defines the operations callers can request, the inputs those operations accept, and the results or errors they return. The contract describes how to interact with the service without knowing how the service is built internally.
Public cloud is a deployment model: a provider operates infrastructure and offers computing resources to a broad customer base. Customers build workloads—their applications and supporting data—using services such as virtual machines, storage, and databases. Public cloud is not one particular product or application design.
A workload is an application or computing task together with the resources it needs. In an on-premises deployment, workloads run on infrastructure in an organization’s own facilities or leased data-center space. The equipment can be owned or leased; the organization does not have to own the building.
Infrastructure is the underlying equipment and facilities that run applications: servers, storage, networking, and the data center around them. The main distinction here is who operates that foundation.
Hybrid cloud connects on-premises systems or a private cloud with public cloud resources so that applications and data flows work across the environments. The useful starting point is the work they share: which application needs which data, and where that work must happen.
A public cloud provider supplies cloud infrastructure. A cloud-only deployment runs the relevant application and its supporting systems in that cloud, without depending on an integrated local application environment.
On-premises infrastructure runs workloads in facilities an organization controls, such as its data center. An exclusively on-premises deployment keeps the relevant workloads there without integrating them with cloud resources.
AWS Outposts brings AWS-designed, AWS-managed hardware into your data center, colocation space, or other on-premises facility. Supported AWS services run on that hardware, close to your local applications and data.
A hybrid deployment connects an on-premises environment with cloud services. That description alone does not tell you what must remain local. The important distinction here is between bringing AWS infrastructure to your facility and connecting your existing infrastructure to AWS storage.
Define the AWS global infrastructure9 lessons
An Availability Zone consists of one or more discrete data centers within an AWS Region. Each zone has redundant power, networking, and connectivity. Zones are physically separated and designed with independent supporting infrastructure.
Failover transfers work from an unavailable or unhealthy resource to another resource that can do the same job. For a web application, this can mean sending new requests to another application instance rather than continuing to use a failed one.
An AWS Region is a separate geographic area containing multiple Availability Zones. If a disruption affects an entire Region, moving work to another Zone within that Region is not enough. Cross-Region disaster recovery prepares a way to restore service in a separate Region.
AWS Elastic Disaster Recovery, or AWS DRS, continuously copies server disk changes to a recovery location. For protection against a source Region outage, choose a different AWS Region as the recovery destination. Another Availability Zone in the source Region does not provide the same separation from a Region-wide outage.
A cache keeps reusable data closer to the people or applications reading it. Reading that copy can take less time than fetching the same data repeatedly from a remote source. The source normally remains authoritative: adding a cache does not replace it.
Fault isolation limits how far a failure can spread. A failure domain is a set of resources that one event can affect together. The scope of the resulting impact is its blast radius.
Resiliency means handling disruption without unacceptable loss of service or data. One way to improve it is to spread resources across failure domains: boundaries within which one failure can affect several resources.
Data sovereignty concerns the legal authority that applies to data and its handling. Physical location matters, but obligations can also depend on who controls the data, whose information it contains, and where people access it. Work with legal and compliance teams to establish requirements before selecting infrastructure locations.
Backup and restore preserves recoverable states of supported resources and uses them to recover data. Disaster recovery has a wider goal: restoring a usable service after a serious disruption. Backup and restore can be part of a disaster recovery strategy; they are not competing alternatives.
Identify AWS compute services20 lessons
Compute is the capacity that executes application instructions, using resources such as CPU and memory. Choosing compute starts with two questions: does the application need a continuously running environment or separate executions, and how much infrastructure must you control?
An Amazon EC2 instance is a virtual server. Its instance type determines its mix of processing power, memory, storage capabilities, and networking capacity. Processing power runs instructions; memory, or RAM, holds data that applications are actively using.
Amazon Lightsail simplifies hosting small websites and applications on virtual servers. A virtual server, also called an instance, gives your software computing resources without requiring you to own a physical machine. Lightsail brings common setup choices into a simplified console.
Amazon Lightsail and Amazon EC2 both let you run applications on virtual servers in AWS. A virtual server provides computing resources without requiring you to buy physical hardware. The main choice between these services is not whether they can host an application, but how much infrastructure you want to configure.
AWS Fargate and AWS Lambda both offer serverless compute: AWS manages the underlying servers. The main difference for this comparison is what your application runs as—a containerized application or an invoked function.
AWS Batch manages batch computing: submitting work that runs without someone interacting with it while it executes. Examples include converting a collection of images, processing data files, and running scientific calculations.
AWS Batch organizes work as jobs. A batch job performs a defined task without someone interacting with it while it runs, such as calculating simulation results or processing a collection of files. Jobs enter a job queue, where they wait to be scheduled onto suitable compute resources.
An Amazon EC2 instance is a virtual server. Its instance type defines its hardware resources, including CPU, memory, and storage and network capabilities. Instance families offer different resource balances: choosing a family means matching those resources to the work an application does.
A container image packages an application, its dependencies, and instructions for starting it. A container is a running instance of that image. Typical containers share their host's operating-system kernel rather than each booting a separate operating system.
Amazon Elastic Container Service (Amazon ECS) is a managed container orchestration service. A container packages an application with the software it needs. ECS coordinates launching and stopping containers so you do not have to build your own container scheduling system.
A container packages an application with the software it needs to run. Kubernetes coordinates containerized applications: it places workloads, maintains the requested number of copies, and replaces failed workloads. Its smallest deployable unit is a Pod, which contains one or more containers.
A container packages application code with the software it needs to run. Containers still need processing power and memory on a server. AWS Fargate supplies that compute capacity without making you provision or maintain the underlying servers.
A container packages an application with the software it needs to run. Container orchestration coordinates running containers: placing them on available computing capacity, replacing failed containers, and adjusting how many copies run. Both Amazon ECS and Amazon EKS provide managed container orchestration.
A container image is a packaged application and the files it needs to run. A container is a running instance of an image. Keeping the package available and managing running instances are different jobs.
A container image packages application code and the files it needs to run. A running container is created from an image. Amazon Elastic Container Registry (Amazon ECR) stores and shares these images without requiring you to operate your own registry infrastructure.
Serverless means the cloud provider operates the servers that run your application. Servers still exist, but AWS handles provisioning, maintaining, and replacing the underlying hosts. You work with resources such as functions or container tasks instead of administering individual machines.
Horizontal scaling changes the number of instances doing the same kind of work. An instance is one running copy of the application or worker. Adding instances is scaling out; removing them is scaling in. The instances work in parallel rather than sending every task through one instance after another.
Amazon EC2 Auto Scaling manages groups of EC2 instances. An Auto Scaling group treats those instances as one unit, allowing the service to launch or terminate instances as capacity needs change.
A load balancer receives incoming application traffic at a shared endpoint and distributes it across application resources called targets. Clients use that endpoint instead of choosing an individual server.
A busy website faces two different problems: repeated requests for the same content, and application work that several servers must share. A content delivery network (CDN) and a load balancer address these different problems.
Identify AWS database services17 lessons
A database organizes information so an application can retrieve and change it reliably. Start the choice with its access patterns: which records it reads, how it finds them, and which changes must happen together. Storing customer information does not, by itself, tell you which database to choose.
Amazon Aurora is a managed relational database engine: software that stores organized data and processes requests for it. Relational applications store data in tables, with rows for individual records and columns for their attributes. For example, a Customers table might contain a customer ID and name for each customer.
Amazon RDS is a managed service for running relational databases. A database engine is the software that stores the data and processes database queries. Choosing the service and choosing the engine are related decisions, but they are not the same decision.
Amazon DynamoDB is a fully managed NoSQL database for key-value and document data. AWS operates the database infrastructure, including server maintenance, software patching, and data replication. You still choose how to organize your data and how your application accesses it.
Amazon DocumentDB and Amazon DynamoDB are managed NoSQL databases. Both can store flexible, nested data, such as a product with a list of sizes and a set of specifications. Seeing document-shaped data is not enough to choose between them.
Amazon ElastiCache is a fully managed, in-memory caching service. A cache holds data that an application can reuse, such as a frequently requested product description. Keeping that data in memory helps reduce latency, the time an application waits for a response.
Amazon Neptune is a fully managed graph database service. A graph represents things and the relationships between them, making connections part of the stored data. AWS manages the underlying database infrastructure so developers can focus on their data and applications.
A database migration moves a database from a source, where it starts, to a target, where it will be used. The database engine is the software that stores, manages, and queries the data. PostgreSQL, MySQL, and Oracle Database are examples of database engines.
A homogeneous migration moves a database while keeping the same database engine. For example, moving a self-managed PostgreSQL database to Amazon RDS for PostgreSQL retains PostgreSQL. The hosting location, operating environment, and database version can change without changing this classification.
A heterogeneous migration moves a database to a different database engine, such as Oracle to PostgreSQL. The new engine may store the same business information but use different rules for defining it and working with it. Copying rows is therefore not enough.
The relational data model organizes information into tables, such as Customers and Orders. It describes how information is structured, independently of the servers or storage running the database. A table's defined structure is called its schema.
The key-value data model stores records as associations between keys and values. A key identifies one record within a collection; its value contains the application data. For example, cart:customer-482 could identify a value containing product identifiers and quantities. The application uses the key to find that particular cart.
A document is a record containing named fields and their values. In {"productId":"P1","name":"Shirt","price":25}, the field name holds text and price holds a number. Documents can also contain true-or-false values.
A data model describes how a database represents information and lets applications work with it. The key-value and document models differ mainly in what database operations focus on—not simply in how the stored data looks.
Amazon DocumentDB is a managed NoSQL document database. Instead of requiring every record to fit the same set of table columns, it stores records as JSON-like documents. A collection holds documents, such as the products in a catalog.
A graph data model represents entities as nodes, also called vertices, and relationships as edges. A person, a product, or an account can be a node. An edge connects two nodes and describes how they are related.
An in-memory cache keeps reusable data in RAM, the computer's working memory. For example, an application can reuse a cached product description instead of asking its database for the same description on every request. This can reduce response time and database work.
Identify AWS network services20 lessons
Amazon VPC gives you a logically isolated virtual network in an AWS Region. You choose its IP address ranges and divide those ranges into subnets, where resources such as EC2 instances receive private IP addresses.
Amazon Route 53 is a managed Domain Name System (DNS) service. DNS helps software turn a readable domain name, such as www.example.com, into information it can use to reach an application. People can keep using the same name even when the application's address changes.
Amazon VPC provides a logically isolated virtual network for AWS resources. You choose IP address ranges, organize resources into subnets, and configure routes and network security controls. A VPC belongs to one AWS Region and can contain subnets in multiple Availability Zones.
A content delivery network (CDN) keeps copies of responses at distributed edge locations. When an edge can reuse a stored response, it answers without fetching that content from its original source, the origin. This can shorten delivery time and reduce repeated origin work.
AWS PrivateLink provides private connectivity to supported services without sending traffic over the public internet. An application in your virtual private cloud (VPC) can use a service without needing a public IP address or an internet gateway for that connection.
A route table tells AWS where to send traffic leaving a subnet. Each route pairs a destination, an IP address range, with a target, such as a gateway. AWS compares the packet’s destination IP address with these ranges to choose a path.
A private application may need to download updates from an internet server. Network Address Translation (NAT) lets it make that outbound IPv4 connection without giving the application its own public IPv4 address.
IP addressing gives a network interface an address that identifies a network endpoint. For example, an EC2 instance's interface might have the private IPv4 address 10.20.1.15. Packets carry source and destination addresses: where they came from and where they are going.
An IP packet carries a source address, identifying its sender, and a destination address, identifying its intended receiver. IP routing examines the destination and selects where to forward the packet next. A route describes a destination range and a forwarding target.
A network filter decides whether traffic may pass. Stateless filtering checks each packet independently, using information such as addresses, protocol, and ports. A reply must match a rule just as the original request must.
Network segmentation separates workloads so that each can reach only what it needs. For example, internet users may need a web server, while only that web server needs the database. These boundaries help limit how far a compromised workload can reach.
A vulnerability is a weakness that could be exploited. For example, a web server running on an Amazon EC2 instance might use a software package with a known security flaw. Vulnerability assessment examines workloads for weaknesses and reports evidence that helps you address them.
DNS, the Domain Name System, helps clients find resources by name. In https://app.example.com/orders, DNS resolves the hostname, app.example.com. It does not interpret the HTTPS scheme or the /orders path.
A virtual private network (VPN) creates a logical connection across an existing network. The connection is called a tunnel: it lets private-network traffic cross another network without requiring a new, dedicated physical circuit between the endpoints.
AWS Site-to-Site VPN connects an on-premises network, such as an office or data center, to a network in AWS. An Amazon VPC is a logically isolated network in AWS where you can run resources.
AWS Client VPN gives remote users secure network access from their devices. A VPN client establishes an encrypted tunnel over the internet to a managed Client VPN endpoint. AWS operates the VPN service infrastructure; you configure who can connect and what they can reach.
A virtual private network (VPN) carries traffic through an encrypted connection. AWS Client VPN and AWS Site-to-Site VPN both provide VPN connectivity, but they connect different things.
AWS Direct Connect connects an organization's network to AWS without sending that traffic over the public internet. Avoiding the changing conditions of internet routes can provide a more consistent network experience, though it does not guarantee a particular application speed.
AWS Direct Connect and AWS Site-to-Site VPN both connect an organization's network, such as a data center or office, to AWS. Their main difference is the network path that carries the traffic.
AWS Transit Gateway is a managed network routing hub. It connects VPCs, which are isolated networks in AWS, and on-premises networks, such as an office or data center. This is called a hub-and-spoke design: each network connects to the hub instead of needing a separate connection to every other network.
Identify AWS storage services29 lessons
A storage model describes how an application sees and accesses data. Object storage stores data as objects, each identified by a key within a bucket. Applications use API requests to store and retrieve them. Amazon S3 provides object storage for data such as images, logs, and backups; it does not present a disk to an operating system.
An S3 storage class determines how an object is stored and charged for. Classes differ in access frequency, retrieval speed, resilience to failures, and cost. A lower storage price does not always mean a lower total bill.
Amazon S3 is an object storage service. It stores data as objects in buckets. A storage class is a choice within that service that determines how an object is stored, with associated access behavior and charges.
S3 Lifecycle rules tell Amazon S3 how to manage objects as they age. For example, you can keep recent logs readily available, move older logs to an archival storage class, and remove logs you no longer need. S3 performs the configured actions automatically, without you handling each object.
Amazon Elastic Block Store (Amazon EBS) provides volumes that an EC2 instance uses like disks. With block storage, the operating system reads and writes blocks of data. It can organize those blocks into a file system to store operating-system files, application files, or database data.
Amazon S3 provides object storage. An object holds data, such as a photo or backup file, along with information describing it. Applications send requests to S3 to store or retrieve individual objects. Each object can be accessed independently; an application does not need an attached disk to retrieve it.
An Amazon EC2 instance is a virtual server. Both Amazon Elastic Block Store (Amazon EBS) and EC2 instance store provide block storage: storage that the operating system can use like a disk. The important difference is whether that storage exists independently of the instance.
Amazon Elastic File System (Amazon EFS) is managed file storage for Linux workloads. It stores files in directories, giving applications a familiar way to organize and use data. AWS manages the storage infrastructure behind the file system.
Amazon FSx provides fully managed file systems: storage organized as files and folders that applications access over a network. Multiple clients can use the same file system, subject to its permissions. This is shared file storage, rather than a separate copy of the files on every application server.
Shared file storage lets several computers work with files and folders in the same file system. Both Amazon EFS and Amazon FSx provide managed file storage: AWS operates the underlying storage infrastructure, rather than requiring you to run your own file servers.
Amazon S3 File Gateway is a type of AWS Storage Gateway that presents file shares backed by Amazon S3. Applications access a share using NFS or SMB, familiar network file-sharing protocols. Each share uses one of these protocols.
Amazon S3 is object storage. An application can store and retrieve objects directly using the S3 API, often through an AWS software development kit. S3 itself does not present a traditional network file share.
Object storage stores data as separate objects. Each object has the data itself, metadata describing it, and a key that identifies it within a bucket. Amazon S3 is an AWS object storage service; an S3 bucket holds objects.
Data archiving preserves information that is no longer needed for everyday work but still has value or must be retained. Completed project files, historical logs, financial statements, and original media files are common examples. Inactive does not mean disposable.
A data lake retains diverse datasets for analysis without requiring everything to fit one relational model before it arrives. It can hold database exports, JSON events, documents, images, and audio. Amazon S3 is a common storage foundation: it stores data as objects in buckets, with each object identified by a key.
Data durability answers, “Will the data still exist when it is needed?” It concerns permanent data loss. Availability answers a different question: “Can I access the data now?” A temporary interruption can make data unavailable without destroying it.
Retrieval latency is the wait between requesting stored data and that data becoming available to read. When choosing storage, start with a simple question: how long can the person or application wait?
S3 Glacier Instant Retrieval, S3 Glacier Flexible Retrieval, and S3 Glacier Deep Archive store rarely accessed data. Their main difference is how quickly that data becomes available when you need it. Instant Retrieval provides millisecond access: you can read an object directly without restoring it first.
An Amazon S3 storage class is a choice of storage pricing and access characteristics for an object. S3 generally bills for the data you store, rather than an empty capacity reservation. However, some classes have minimum billable object sizes, so very small objects can be billed as larger ones.
Block storage exposes numbered chunks of storage called blocks. The operating system sees a block device, much like a disk. It can read and overwrite particular blocks without replacing the entire disk.
An Amazon EC2 instance is a virtual server. It can use block storage: storage presented to its operating system as a disk made of addressable blocks. The operating system can put a file system on that disk to store files, just as it would on a physical server.
Ephemeral storage is storage whose data should be treated as expendable. The key question is not what kind of disk it uses, but which events cause its data to disappear.
A shared file system gives several servers access to the same files and directories over a network. A server mounts the filesystem or connects to a share, then applications use file paths to open, read, and write files. NFS is common with Linux and Unix; SMB is common with Windows.
Hybrid cloud storage coordinates storage in an organization’s local environment with storage in a public cloud. A storage gateway connects the two: applications use a familiar file share, while the gateway handles cloud storage access.
A retention period describes how long data should be kept. A retention policy identifies the data, why it is needed, and the event that starts the clock. That event could be creation of a record or closure of a business transaction. Retention periods come from business, legal, and recovery requirements—not one universal AWS default.
Storage tiering means placing data in a storage class that matches how often it is read and how quickly it must be available. Frequently read data usually needs economical, immediate access. Rarely read data may benefit from a lower storage price. These needs can change as data gets older.
Data retention answers how long data should be kept. Storage tiering answers which storage class should hold it while it is kept. These are separate decisions about the same data.
A backup preserves a recoverable state of data. That saved state is called a recovery point. If someone deletes records or a software error corrupts files, a recovery point from before the damage can provide the earlier data. A restore uses that recovery point to recover the data.
AWS Backup is a fully managed service for centrally managing backups across supported resources. Instead of maintaining separate backup scripts for each resource, you define reusable rules in a backup plan. AWS Backup uses those rules to automate backup work.
Identify AWS artificial intelligence and machine learning (AI/ML) services and analytics services26 lessons
Customer reviews, emails, and articles contain unstructured text: people write freely rather than filling in predictable fields. The information is there, but an application cannot simply read a column named “customer opinion” to find it.
Amazon Comprehend analyzes text to uncover information in it. Choose it when an application needs findings about written content, such as the sentiment of customer feedback or the people and places mentioned in documents. The text is material to study; the result is information the application can use.
Amazon Polly is a managed text-to-speech (TTS) service: it converts written text into spoken audio. This conversion is called speech synthesis. AWS runs the speech-generation technology, so developers do not need to build or train their own speech model.
Amazon Polly turns supplied text into spoken audio. This is called text-to-speech, or speech synthesis. The application supplies the words; Polly produces speech that the application can play. Reading an article or announcing a delivery update does not require understanding a user's request.
Amazon Rekognition is a managed AWS service that analyzes images and videos. Applications call its APIs to detect visual content, such as objects, scenes, faces, text, and potentially inappropriate content. Developers do not need to build the underlying computer vision system.
Amazon SageMaker is a unified platform for data, analytics, and AI. It brings together capabilities for working with data, analyzing it, and developing AI applications. SageMaker Unified Studio provides a shared development environment for these activities.
Amazon Q is in maintenance since June 30, 2026 (2026-06-30): it is closed to new customers, while existing customers can keep using it. This lesson describes the service as AWS offered it, not as a recommendation for new workloads. AWS's migration guide is https://docs.aws.amazon.com/amazonq/latest/qbusiness-ug/qbusiness-availability-change.html.
Amazon Athena is a serverless query service. You write SQL to analyze data directly in Amazon S3, without first loading it into a separate database. AWS manages the computing resources that run the query.
Amazon Redshift is a managed data warehouse: a service designed to store and analyze large datasets. A warehouse brings data together for questions such as “How much did each region sell this year?” rather than focusing on individual actions such as updating one customer's address.
Amazon Athena and Amazon Redshift both let you analyze data with SQL, a language for asking questions about data. The important difference is not the language: it is whether you want to query files where they already sit or use a data warehouse for analytics.
Amazon EMR is a managed service for processing large amounts of data using frameworks such as Apache Spark and Apache Hadoop. A framework provides software for organizing and running data-processing work; EMR helps set up and operate the resources that run it.
Amazon EMR runs data-processing applications using frameworks such as Apache Spark and Hadoop. A framework provides tools for dividing work across computing resources. Choose EMR when you need to build or adapt how a large processing job runs, rather than just submit a query.
AWS Glue helps prepare and integrate data. Its managed ETL jobs extract data from sources, transform it, and load the result into a destination. Choose this approach when you need to clean inconsistent values, combine sources, or produce a reusable dataset.
Amazon OpenSearch Service is a managed search and analytics service. It helps applications find matching information quickly and helps people investigate logs. AWS manages much of the underlying infrastructure, while you decide what data to send and how to search it.
Amazon QuickSight is a cloud-based business intelligence service: it helps people turn data into charts, tables, and interactive dashboards. For example, a sales team can explore revenue by month or region instead of reading rows of sales records.
Amazon Lex helps you build conversational interfaces that accept voice or text. For voice input, speech recognition turns speech into text. Natural language understanding interprets what the user means. The user's spoken or typed input is called an utterance.
A scanned document is a picture of a page. A person can read the words, but an application needs those words as data to search, copy, or process them. Amazon Textract is a managed machine learning service that extracts text and data from documents, reducing manual retyping. You do not need to build or train your own recognition model.
Amazon Transcribe is a managed automatic speech recognition service: it recognizes spoken words in audio and returns text. AWS runs the speech recognition models, so you do not need to train or operate your own model.
Amazon Translate is a managed service that converts text from one language into another. The source language is the language of the original text; the target language is the language you want to produce.
Amazon Lex helps you build an interactive conversational bot using voice or text. Choose it when an application needs to understand a user's request and guide the user through a conversation, such as a bot that helps arrange an appointment.
Amazon Textract and Amazon Transcribe both help turn information into text that software can use. The main difference is their input: Textract extracts content from documents, while Transcribe converts speech into text.
Amazon Transcribe turns speech into text. This is called transcription: spoken words become a written transcript in the language being spoken. For example, an English recording becomes English text that someone can read instead of listening to the recording.
AWS Glue is a managed data integration service that helps prepare and combine data for analytics. Its ETL jobs extract data from a source, transform it into a useful form, and load the results into a destination.
Streaming data arrives continuously as records, such as website clicks, application logs, or sensor readings. Amazon Data Firehose is a fully managed delivery service: it receives those records, temporarily collects them, and delivers them to a configured destination.
Amazon Kinesis Data Streams is a managed service that captures and retains a continuous flow of data records. Applications can process records as they arrive instead of waiting for a large batch to accumulate.
Streaming data is a continuing flow of records, such as application events. Amazon Data Firehose and Amazon Kinesis Data Streams both handle streaming data, but they solve different problems.
Identify services from other in-scope AWS service categories34 lessons
An event is a record of something that happened, such as an instance stopping or an order being placed. Amazon EventBridge is a serverless service that receives events and routes them to destinations. You do not manage servers for its event buses.
Amazon Simple Notification Service (Amazon SNS) distributes messages from senders to interested applications and people. It uses a publish/subscribe model: senders publish messages, and receivers subscribe to receive them.
Amazon Simple Queue Service (Amazon SQS) is a fully managed message queuing service. A producer sends a message describing work to a durable queue. A consumer retrieves that message and does the work. AWS operates the queue infrastructure; you write the applications that use it.
An event describes something that happened, such as an order being placed. Choose Amazon EventBridge when the main job is to examine application events and route matching events to targets. An event bus receives events, and rules select where they go.
An event describes something that happened, such as an order being placed. Amazon EventBridge receives events on an event bus and uses rules to route matching events to configured targets. Choose it when the main question is: which targets should receive this event?
Amazon Simple Notification Service (Amazon SNS) distributes notifications. A publisher sends a message to a topic, a named destination with subscribers. SNS pushes the message to subscribed destinations. Sending one published message to multiple subscribers is called fan-out.
Amazon Connect is an AWS-managed cloud contact center. It helps a business handle customer voice calls and chats, guide customers through automated steps, and route interactions to human agents.
Amazon Simple Email Service (Amazon SES) is a managed service for sending email from applications. Your application supplies the recipient, sender, and message; SES handles the outgoing email infrastructure. You do not install, patch, or operate mail servers.
AWS Support plans determine how you get help with AWS and how much ongoing guidance you receive. Basic Support is included for all AWS customers. It provides account and billing assistance, while technical learning and troubleshooting rely on self-service resources such as AWS documentation and the AWS re:Post community.
AWS Health Dashboard reports AWS operational events, including service disruptions and planned changes. It helps you understand what AWS reports and whether the event is relevant to your resources. It does not measure your application's response time or check that your application works correctly.
AWS Health Dashboard helps answer, “What AWS operational events could affect my environment?” It presents information about AWS service disruptions, scheduled changes, and events relevant to your account. This information helps you understand an event’s impact and prepare or respond.
AWS CodeBuild is a managed service that runs commands to compile source code, run tests, and package software. The resulting files are called build artifacts: for example, an application ZIP file that can later be deployed.
AWS CodePipeline is a managed service that coordinates the steps used to release a software change. You define a pipeline: a repeatable workflow that takes a source revision through the work needed before deployment. This reduces manual handoffs and makes the release process consistent.
A software release involves different jobs: preparing source code, building it, checking it, and deploying it. AWS CodeBuild and AWS CodePipeline address different parts of that work rather than being interchangeable services.
A single application request can pass through several services before returning a result. AWS X-Ray helps you understand that journey through distributed tracing: connecting records of work across application components.
Amazon AppStream 2.0 is a fully managed application streaming service. It lets users work with desktop applications through a supported web browser without installing those applications on their own devices.
Amazon WorkSpaces is a managed cloud desktop service. The desktop operating system and its applications run in AWS, rather than on the user's laptop or tablet. This lets an organization provide a work desktop without putting that desktop's computing hardware at the user's location.
Amazon WorkSpaces Web is now called Amazon WorkSpaces Secure Browser. It provides a managed, isolated cloud browser for accessing internal websites and SaaS applications—software used through the web.
Amazon AppStream 2.0 and Amazon WorkSpaces both let people interact with software running in AWS. The user's device displays the remote output and sends input, such as keyboard actions and mouse clicks. The device is not doing the application's main computing work.
Amazon WorkSpaces provides a complete cloud desktop: a desktop environment running in AWS where users can work with applications, files, and websites. Their device displays the remote desktop and sends their interactions to it.
Amazon WorkSpaces Web, now called Amazon WorkSpaces Secure Browser, provides a managed browser running in AWS. Users access internal websites and software-as-a-service (SaaS) applications through that remote browser.
A frontend is the part of an application people interact with: pages in a web browser or screens in a mobile app. A backend provides capabilities such as user sign-in, access to shared data, and storage.
AWS IoT Core is a managed service that lets connected devices exchange messages with cloud applications. A sensor can send a temperature reading, and an application can send a command back. AWS operates the messaging infrastructure, so you do not install, patch, or scale message-broker servers.
An IoT device is a connected physical device, such as a temperature sensor or a smart thermostat. A fleet is a collection of these devices. Running an IoT system means both exchanging messages with devices and keeping the fleet organized and maintained.
An event is a record that something happened, such as OrderPlaced. A producer publishes that fact, and consumers react to it. Fulfillment and reporting can respond to the same order without the order application directly calling each of them.
A message queue holds messages describing work until an application can process them. The application creating the work is the producer; the application performing it is the consumer, often called a worker.
Publish-subscribe messaging separates sending a message from responding to it. A publisher sends a message to a named destination called a topic, rather than contacting every recipient directly.
Messaging lets one application send information without waiting for another application to finish acting on it. The key choice is whether recipients should share the work or each receive a copy.
AWS Support plans define the assistance you can receive from AWS. They do not change the resources running your application. Buying a higher tier does not automatically make an application more available or transfer your operational responsibilities to AWS.
AWS Support plans provide an ongoing support relationship for customers using AWS. They help teams obtain assistance with account questions and, depending on the plan, technical troubleshooting and guidance. The purpose is to support the customer's use of AWS, rather than deliver a separately defined cloud transformation project.
AWS Professional Services and AWS Solutions Architects both help customers succeed on AWS. They provide assistance through people, rather than cloud resources you create in the AWS console. The main distinction is scoped project delivery versus architectural guidance.
Continuous integration (CI) means developers frequently merge small changes into shared source code and get automated build and test feedback. This catches problems early; an occasional automated build alone is not effective CI if changes stay isolated for long periods.
Distributed tracing follows one request across the services that handle it. An order request might enter an order service, which calls an inventory service. Both services could appear healthy overall while that particular request is slow or fails.
AWS IoT Device Management helps operators organize and remotely manage a fleet of connected devices. A registered device is represented by a thing. A thing group collects these representations so operators can target a useful set of devices instead of selecting each device individually.
Compare AWS pricing models1 free13 lessons
Usage-based pricing connects charges to measured quantities, such as compute running time, stored data, or requests. With an option such as EC2 On-Demand, you pay for billable usage without accepting a long-term purchasing commitment. Measured usage is not the same as useful work: a running instance can incur compute charges even when its application is idle.
A commitment discount lowers the price of eligible usage in exchange for an obligation over a term. The important question is what usage the agreement can cover. Two common approaches are resource-based and spending-based commitments.
An Amazon EC2 instance is a virtual server. With On-Demand Instances, you pay for running instance usage without a long-term commitment. Running is not the same as doing useful work: an instance with an idle CPU still incurs compute charges. Stopping an EBS-backed On-Demand instance generally stops its instance compute charges, but retained storage and other resources can still cost money.
An EC2 instance is a virtual server. Purchasing options change how you pay for it, whether spare capacity can be reclaimed, whether capacity is reserved for you, and whether its physical hardware is dedicated to your use.
Savings Plans is a pricing model, not a resource you deploy. You commit to a spending amount in dollars per hour for a one-year or three-year term. AWS automatically applies discounted rates to eligible usage that matches your plan.
An EC2 Reserved Instance (RI) is a billing benefit, not a separate virtual machine. In return for a term commitment, it discounts eligible EC2 usage automatically. You do not attach it to a particular running instance. Unused benefit is not saved for later.
Workload predictability means confidence in future compute needs. It does not mean demand must stay constant. Start with measured usage over a period that includes ordinary busy and quiet times, relevant business cycles, and seasonal changes. Historical measurements are evidence, not a promise that future demand will be identical.
Interruption tolerance means work can still finish correctly even if the computer doing it stops. Amazon EC2 Spot Instances use spare AWS compute capacity at discounted prices, but AWS can reclaim that capacity. They suit work that can pause, restart, or move to another worker.
Workload predictability means confidence in estimating future compute demand: how much computing power you will need and when you will need it. Predictable does not mean constant. A job that runs every evening can have predictable demand even though it is idle during the day.
Data egress means data leaving a defined boundary. A file leaving an AWS Region is egress from that Region, even if it goes to another AWS Region. Ingress is data entering a boundary. Always name the boundary before describing the direction.
Data transfer means moving bytes between locations. Direction matters: data arriving in AWS from the internet is generally free of data-transfer charges. Data leaving AWS for the internet is generally charged, subject to allowances and service-specific exceptions. Free transfer does not mean free storage, requests, or processing.
Data locality means putting data near the applications or people that use it, or running computation near its data. Moving less data across chargeable boundaries can reduce transfer costs. A small request can return a large dataset, so follow the direction and volume of the actual data.
A storage bill can include both capacity, the amount of storage, and performance, how quickly it can handle work. Paying for storage does not always mean paying only for the bytes your application has written.
Understand resources for billing, budget, and cost management13 lessons
AWS Billing and Cost Management is where you inspect AWS charges and billing documents. Open its console, choose Bills, and select the billing month you want to review. Check the month first so that you do not confuse current charges with an earlier bill.
AWS Support provides account and billing assistance to customers on every support plan, including Basic Support. You do not need to purchase a paid support plan to ask about an invoice, a payment problem, or an account-related issue.
AWS Support Center is where you create and manage support cases: recorded conversations with AWS about an account, a bill, or an eligible technical problem. It is not a monitoring system or a cost-analysis dashboard. Open it from the AWS Management Console while signed in to the relevant account with permission to manage support cases.
Usage-based pricing connects a charge to a measurable quantity, such as compute time, stored data, requests, or data transferred. Start with a defined period, such as a month, and estimate each chargeable quantity separately. A running resource can generate charges even when it is not doing useful application work.
A cost allocation tag lets you organize eligible AWS charges by a business label, such as department or project. A tag has a key and a value: in CostCenter=CC104, the key is CostCenter and the value is CC104. Billing tools can use these values to show costs for each cost center.
A cost allocation tag is a label used to attribute AWS spending to a project, team, or other category. A tag has a key and a value: in Project=Atlas, the key is Project and the value is Atlas.
Cost allocation means attributing charges to the teams, products, or departments responsible for them. It answers “Where does this cost belong?” rather than “What price does AWS charge?” Allocating a charge does not itself reduce usage or change its price.
The AWS Cost and Usage Report (CUR) contains detailed cost and usage line items. Cost allocation tags help connect eligible costs to labels such as a project or department.
A cost allocation tag, such as Project, lets you organize spending by a business label. For this lesson, assume the tag is already available for cost analysis. Both AWS Cost Explorer and AWS Cost and Usage Report (CUR) support analyzing spending by that tag.
Resource tagging adds descriptive information to a resource, usually as a key and value such as Environment=Development. Tags help people and tools find and manage resources. One resource can have several tags for different purposes.
AWS Budgets monitors costs or usage against a target you define for a period, such as a month. It helps you notice when consumption is moving beyond your plan so you can respond.
AWS Cost Explorer turns AWS cost and usage data into interactive charts and tables. A cost view shows money spent; a usage view shows quantities consumed, such as running hours or storage. Compare usage with matching units: adding unrelated quantities does not produce a meaningful total.
AWS Budgets monitors costs against a budget you define. Choose it when your question is, “Has spending crossed the threshold I want to watch?” Its value is ongoing attention to a defined target, with notifications configured for that target.
Identify AWS technical resources and AWS Support options6 lessons
AWS Prescriptive Guidance and AWS Knowledge Center are official AWS technical resources, not services that deploy or manage your infrastructure. The useful distinction is the kind of help you need: a recommended approach for achieving an objective, or a focused answer to a question or problem.
AWS re:Post is a public technical knowledge community at https://repost.aws/. It contains questions, answers, and technical articles about building and operating on AWS. Community members and AWS experts share knowledge so that others can reuse it.
The AWS Acceptable Use Policy (AUP) sets boundaries for how customers and their users may use AWS services. These boundaries apply to activity through cloud resources, including virtual machines, websites, storage, and messaging systems. The AUP works alongside service agreements and applicable law to protect customers, internet users, and AWS infrastructure.
The AWS Partner Network (APN) is an ecosystem of participating organizations. It helps those organizations develop AWS skills and grow their businesses. It is not a cloud runtime service or an AWS Support plan.
AWS Marketplace helps customers purchase third-party software. The purchase agreement defines the customer's license entitlements: the rights to use that software. These can include a product edition, enabled features, a licensed quantity, and a period of use. An entitlement record represents those rights; it does not replace the agreement.
AWS Solutions Architects are customer-facing technical advisors who help organizations design and improve workloads on AWS. A workload is an application or system that delivers a business function. Architectural guidance connects what that workload needs to possible designs.