1 What Is Cloud Computing? · AWS Cloud Practitioner

In this chapter you will learn what cloud computing is and how it differs from running your own data center, the three service models (IaaS, PaaS, SaaS) and who manages what in each, the cloud, hybrid, and on-premises deployment models, the six advantages of cloud computing that AWS describes, the difference between scalability and elasticity and between high availability and fault tolerance, and the four main disaster recovery strategies. This chapter maps to Domain 1, Cloud Concepts, Task Statement 1.1 (define the benefits of the AWS Cloud) and supports Task Statement 1.2 (identify design principles of the AWS Cloud).

1.1 A working definition

AWS defines cloud computing as the on-demand delivery of IT resources over the internet with pay-as-you-go pricing. Every word in that sentence matters, and the exam expects you to recognize each part.

On-demand means you request resources when you need them and get them in minutes or seconds, without a purchase order, a delivery truck, or an installation team.

IT resources means computing power, storage, databases, networking, analytics, machine learning, and hundreds of other services.

Over the internet means you reach those resources through a web console, a command line, or an application programming interface (API) instead of walking into a server room.

Pay-as-you-go pricing means you pay for what you actually use, typically by the second, hour, gigabyte, or request, and you stop paying when you stop using it.

Think of the difference between owning a car and using a ride service. When you own a car, you pay for it up front, pay for insurance and maintenance whether you drive or not, and you are stuck with the size you bought. With a ride service, you pay per trip, choose a bigger vehicle when you have luggage, and never think about oil changes. Cloud computing turns IT from something you own into something you consume.

Good to know

The idea of renting computing is old; mainframe time-sharing existed in the 1960s. What changed in the 2000s was the scale and the self-service model. Amazon Web Services launched its first widely used services, Amazon S3 and Amazon EC2, in 2006.

1.2 On-premises versus the cloud

Before the cloud, most organizations ran their applications on-premises: in their own buildings or in data center space they leased. To launch a new application they had to:

1. Estimate how much capacity they would need, often for three to five years ahead.
2. Get budget approval for a large capital purchase.
3. Order servers, storage, and network equipment, and wait weeks for delivery.
4. Rack, cable, power, and cool the hardware.
5. Install operating systems and software, then patch and maintain them for years.
6. Eventually retire the equipment and start the cycle again.

All of that work is real, and none of it makes the application better for customers. AWS calls this kind of work undifferentiated heavy lifting: effort every company has to spend but that does not set any company apart from its competitors.

In the cloud, AWS owns and operates the physical infrastructure. You still design and run your applications, but you provision resources with a few clicks or a line of code, and you release them just as easily.

Concern On-premises AWS Cloud
Up-front cost Large capital expense None for most services
Time to getaserver Weeks to months Minutes
Capacity planning Guess yearsahead Adjustas you go
Physical security Your responsibility AWS responsibility
Global presence Build or lease sites Choosea Region

Common trap

Moving to the cloud does not remove all of your responsibilities. You still own your data, your user access, and (depending on the service) your operating systems and application code. The shared responsibility model, covered in the security chapters (Chapters 5 to 7), spells out exactly who does what.

1.3 Service models: IaaS, PaaS, and SaaS

Cloud services are grouped into three broad models depending on how much of the stack the provider manages for you. The more the provider manages, the less flexibility you have, but the less work you do.

Who manages each layer?

Figure 1.1: Who manages each layer in on-premises, IaaS, PaaS, and SaaS. Moving right, the provider takes on more of the stack and the customer manages less.

Infrastructure as a Service (IaaS)

IaaS gives you the basic building blocks: virtual servers, storage, and networking. The provider runs the physical data center, the hardware, and the virtualization layer. You manage everything from the operating system upward: patches, runtime, application, and data. IaaS offers the most control and is closest to running your own servers. AWS examples: Amazon EC2 (virtual servers), Amazon EBS (block storage volumes), Amazon VPC (your own isolated network).

Platform as a Service (PaaS)

PaaS removes the need to manage the underlying infrastructure and operating system. You bring your application code and data, and the platform handles deployment, capacity, and patching of the platform itself.

AWS examples: AWS Elastic Beanstalk (upload your code and it provisions and manages the servers, load balancer, and scaling for you), Amazon RDS (a managed relational database where AWS patches the database engine and OS), and AWS Lambda (run code without managing servers).

Software as a Service (SaaS)

SaaS delivers a complete application that you simply use. The provider runs and manages everything. Your responsibility shrinks to your data, your users, and how you configure the application.

Examples: web-based email, online office suites, customer relationship management (CRM) systems. Many third-party SaaS products are sold through the AWS Marketplace, and some AWS services, such as Amazon Connect (a cloud contact center), are complete applications that you configure rather than build.

Exam tip

Look for the phrase that signals the model. “Full control over the operating system” points to IaaS. “Developers just upload code” or “no need to manage the underlying infrastructure” points to PaaS. “A finished application used by end users” points to SaaS.

Real-world example

Real-world scenario: A small online bookstore starts by running its website on Amazon EC2 instances that its two engineers patch every month (IaaS). After a missed patch causes an outage, it moves the site to AWS Elastic Beanstalk and the database to Amazon RDS (PaaS), freeing the engineers to work on features. For email and payroll, the company signs up for hosted applications (SaaS) and does not run any servers for them at all.

1.4 Deployment models

Separate from what the provider manages is the question of where your workloads run. The exam uses three deployment models.

Figure 1.2: The three deployment models. A hybrid deployment connects on-premises infrastructure to the cloud over a private connection or a

Vpn.

Cloud (all-in cloud)

All parts of the application run in the cloud. Some applications are built in the cloud from the start (“cloud-native”); others are migrated from on-premises. A cloud deployment can use low-level infrastructure such as EC2 or higher-level managed services such as Lambda and Amazon DynamoDB.

Hybrid

Some resources run on-premises and some run in the cloud, connected so they work together. Organizations choose hybrid deployments for many reasons: legacy applications that are hard to move, data that regulations say must stay in a particular location, or a gradual migration in progress. A common example is keeping a core database on-premises while running new web front ends in AWS. AWS services that support hybrid designs include AWS Direct Connect (a dedicated private network link), AWS Site-to-Site VPN, AWS Storage Gateway, and AWS Outposts (AWS-managed racks installed in your own data center).

On-premises (private cloud)

Resources are deployed in the organization’s own data center, often using virtualization and resource management tools to provide some cloud-like self-service. This is sometimes called a private cloud. It offers maximum control, but it does not deliver most of the economic benefits of the public cloud, because the organization still buys and owns all the capacity.

Exam tip

When a question mentions “connecting on-premises infrastructure to AWS” or “keeping some workloads in the data center”, the answer is hybrid. When the question says “private cloud”, think on-premises.

1.5 The six advantages of cloud computing

AWS describes six advantages of cloud computing. They appear on the exam constantly, often as the answer choices themselves. Learn the exact phrasing.

1. Trade fixed expense for variable expense

Instead of investing heavily in data centers and servers before you know how you will use them, you pay only when you consume computing resources, and only for how much you consume. In accounting terms, you shift spending from capital expense (CapEx) to operating expense (OpEx).

Example: A startup can launch a product with no up-front hardware purchase. If the product fails, it shuts the resources down and stops paying.

2. Benefit from massive economies of scale

Because AWS aggregates usage from hundreds of thousands of customers, it can buy hardware, power, and network capacity more cheaply than any single company. AWS passes those savings on through lower pay-as-you-go prices, and has reduced prices many times over its history.

Example: A mid-sized company could never negotiate the electricity rates or server prices that AWS gets, but it benefits from them through its bill.

3. Stop guessing capacity

On-premises, you must guess how much capacity you need. Guess too high and you pay for idle servers; guess too low and your application slows or fails when demand arrives. In the cloud, you can access as much or as little as you need and scale up or down in minutes.

Example: An online retailer does not need to buy servers in July for the holiday peak in December. It adds capacity when the peak arrives and removes it afterward.

4. Increase speed and agility

New IT resources are only a click away, which reduces the time to make resources available to developers from weeks to minutes. Lower cost and time for experiments means teams can try more ideas.

Example: A data science team can spin up a large cluster for an afternoon to test a hypothesis, then delete it. On-premises, the same test might take months to approve.

5. Stop spending money running and maintaining data centers

Focus on projects that differentiate your business, not on racking, stacking, and powering servers. This is the cure for undifferentiated heavy lifting.

Example: A bank moves its staff away from managing cooling systems and hardware refreshes toward building better mobile banking features.

6. Go global in minutes

Deploy your application in multiple AWS Regions around the world with a few clicks, giving users lower latency and a better experience at minimal cost.

Example: A gaming company based in Canada launches servers in Europe and Asia-Pacific Regions on the same day it decides to expand.

Exam tip

Match the scenario to the advantage. Up-front costs avoided = fixed to variable expense. Lower prices from AWS’s size = economies of scale. Avoiding over- or under-provisioning = stop guessing capacity. Faster experiments = speed and agility. Serving users in other continents = go global in minutes.

1.6 Scalability and elasticity

These two words are related but not identical, and the exam likes to test the difference.

Scalability is the ability of a system to handle a growing amount of work by adding resources. It is about capacity to grow. There are two directions:

Scale up (vertical scaling): make a single resource bigger, for example moving from a small EC2 instance to a larger one with more CPU and memory. There is always a ceiling, and resizing an instance usually requires a brief stop. Scale out (horizontal scaling): add more resources of the same kind, for example going from two web servers to ten behind a load balancer. This has a much higher ceiling and also improves availability.

Elasticity is the ability to automatically add resources when demand rises and remove them when demand falls, so capacity tracks demand closely. Elasticity is scalability that goes in both directions, automatically and quickly. On AWS, Amazon EC2 Auto Scaling and the built-in scaling of services such as AWS Lambda and Amazon DynamoDB provide elasticity.

Elastic capacity stays just above demand: little waste, no shortfall.

Figure 1.3: Fixed on-premises capacity versus elastic cloud capacity. Fixed capacity wastes money when demand is low and fails customers when demand exceeds it; elastic capacity follows the demand curve.

Figure 1.3 shows why this matters. A company with fixed capacity must choose a size. The shaded area above the demand

curve is money spent on idle servers. The area where demand rises above the fixed line represents lost customers and slow pages. Elastic capacity follows the curve in small steps, so there is little waste and little shortfall.

Common trap

If a question emphasizes automatically matching resources to changing demand, especially scaling down to save money, the answer is elasticity, not scalability. Scalability alone describes the ability to grow.

1.7 High availability and fault tolerance

Both terms describe how well a system keeps working when something breaks. The difference is in the user experience during a failure.

High availability (HA) means the system is designed to stay operational with minimal downtime. When a component fails, the system recovers quickly, perhaps with a brief interruption, by shifting work to healthy components. On AWS, a typical HA design runs EC2 instances in two or more Availability Zones (separate data center locations within a Region) behind an Elastic Load Balancer. If one Availability Zone has a problem, the load balancer sends traffic to the others.

Fault tolerance goes further: the system continues operating without interruption even when components fail, because redundant components are already running and taking over seamlessly. Fault tolerance usually costs more, because you pay for duplicate capacity that is always on.

A simple way to remember it: a car with a spare tire in the trunk is highly available, because you can be back on the road in fifteen minutes. A plane with multiple engines is fault tolerant, because it keeps flying if one engine fails. Some AWS services are built to be highly available or fault tolerant by default. For example, Amazon S3 stores data redundantly across multiple Availability Zones, and Amazon DynamoDB replicates data across three Availability Zones in a Region automatically.

Exam tip

“Deploy across multiple Availability Zones” is the standard answer for improving availability. “Deploy across multiple Regions” is usually the answer for disaster recovery from a regional event or for serving global users with low latency.

1.8 Agility

Agility is mentioned in the six advantages, but it is worth its own section because the exam frames it in several ways. In cloud terms, agility means an organization can move faster: provision resources quickly, experiment cheaply, and fail without heavy losses. Agility shows up as:

Speed: resources in minutes instead of weeks.

Experimentation: because nothing is bought up front, a failed experiment costs only the hours used. Innovation: teams have immediate access to advanced services, such as machine learning, analytics, and the Internet of Things (IoT), without building them.

Automation: infrastructure can be described in code and recreated identically, which makes releases faster and safer.

Real-world example

Real-world scenario: A retail chain wants to test whether product recommendations increase sales. Instead of buying servers and hiring a machine learning platform team, it uses managed AWS AI services for a four-week pilot in one Region. The pilot works, so it expands to all stores. Had it failed, the company would have deleted the resources and lost only a few weeks of usage charges.

1.9 Disaster recovery strategies

A disaster is any event that seriously disrupts an application: a flood, a power failure, a ransomware attack, an accidental deletion of production data. Disaster recovery (DR) is the plan for getting back to normal. Two measures define how good a DR plan needs to be:

Recovery Time Objective (RTO): the maximum acceptable time between the disaster and the restoration of service. “We must be back online within 4 hours.”

Recovery Point Objective (RPO): the maximum acceptable amount of data loss, measured in time. “We can lose at most 15 minutes of transactions.”

Lower RTO and RPO mean faster recovery and less data loss, but they also cost more. AWS describes four strategies along this spectrum.

Figure 1.4: The four disaster recovery strategies. Moving right, recovery time and data loss shrink while cost and complexity grow.

Backup and restore

You back up data and configurations to AWS (for example, to Amazon S3 or using AWS Backup), and in a disaster you restore them and rebuild the environment. This is the cheapest option and the slowest. RTO and RPO are typically measured in hours.

Pilot light

The most critical core of the system, usually the database, is kept running and continuously replicated in the recovery Region, but application servers are switched off. In a disaster you “turn up the flame”: launch the servers from prepared images and scale them. RTO and RPO are typically tens of minutes.

Warm standby

A scaled-down but fully functional copy of the whole environment runs at all times in the recovery Region. It can serve traffic immediately at reduced capacity and is scaled up to full size in a disaster. RTO and RPO are typically minutes.

Multi-site active/active

The full workload runs in two or more Regions at the same time, all serving traffic. If one Region fails, the others simply absorb its load. This is the most expensive option and offers near-zero downtime and data loss.

Strategy What runs in recovery site Relative cost Typical RTO
Backup and restore Nothing, only backups Lowest Hours
Pilot light Core data only Low Tens of minutes
Warm standby Small full copy Medium Minutes
Multi-site Full copy, live Highest Near zero

Exam tip

A useful memory trick: in pilot light the data is live and the servers are off; in warm standby the whole system is on, just small. If the scenario says “minimal running resources, core database replicated”, it is pilot light.

Good to know

AWS Elastic Disaster Recovery is a service that continuously replicates servers to AWS and lets you recover them quickly, which is a common way to implement pilot light or warm standby for on-premises and cloud servers.

Chapter 1 summary

Cloud computing is the on-demand delivery of IT resources over the internet with pay-as-you-go pricing. On-premises IT requires up-front capital purchases, long lead times, and capacity guesses; the cloud removes this undifferentiated heavy lifting.

IaaS gives the most control (you manage the OS and up), PaaS lets you focus on code and data, and SaaS delivers a finished application.

Deployment models are cloud, hybrid (on-premises connected to the cloud), and on-premises (private cloud). The six advantages: trade fixed for variable expense, economies of scale, stop guessing capacity, speed and agility, stop running data centers, go global in minutes.

Scalability is the ability to grow; elasticity automatically grows and shrinks capacity with demand. High availability minimizes downtime; fault tolerance continues with no interruption, at higher cost. Disaster recovery strategies from cheapest and slowest to most expensive and fastest: backup and restore, pilot light, warm standby, multi-site active/active.

RTO is how long you can be down; RPO is how much data you can lose.

Key terms
Term Definition
Cloud computing On-demand IT resources over the internet with pay-as-you-go
pricing
IaaS Provider manages hardwareand virtualization; you manage OS
and up
PaaS Provider manages infrastructureand OS; you manage codeand
data
SaaS Provider manages the wholeapplication; you use it
Hybrid deployment On-premises resources connected to cloud resources
CapEx / OpEx Up-front capital spending / ongoing operating spending
Scalability Ability to handle more load by adding resources
Elasticity Automatically adding and removing resourcesas demand changes
High availability Designed for minimal downtime when components fail
Fault tolerance Continues operating with no interruption during failures
RTO Maximum acceptable downtimeaftera disaster
RPO Maximum acceptable dataloss, measured in time
Pilot light DR with core datareplicated and servers off until needed
Warm standby DR with ascaled-down full copy always running
Practice questions

1. A company wants to stop buying servers based on three-year capacity forecasts that are often wrong. Which advantage of cloud computing does this describe?

A. Go global in minutes B. Stop guessing capacity
C. Benefit from massive economies of scale D. Increase speed and agility

2. Which statement BEST describes cloud computing?

A. Renting physical servers in a colocation facility on a five-year contractB. On-demand delivery of IT resources over the internet with pay-as-you-go pricing
C. Running virtual machines in a company’s own data

center

D. Purchasing software licenses for installation on employee

laptops

3. A development team wants to deploy a web application by uploading its code, without managing servers, operating systems, or load balancers. Which cloud service model fits BEST?

A. Infrastructure as a Service (IaaS) B. Platform as a Service (PaaS)
C. Software as a Service (SaaS) D. On-premises virtualization

4. A company must keep its customer database in its own data center for regulatory reasons but wants to run its new web front end on AWS. Which deployment model is this?

A. Cloud B. Private cloud
C. Hybrid D. Multi-site

5. Which AWS service is an example of Infrastructure as a Service?

A. Amazon EC2 B. AWS Elastic Beanstalk
C. A hosted customer relationship management (CRM)

application

D. AWS Lambda

6. An application’s traffic rises sharply every weekday morning and drops overnight. The company wants resources to be added and removed automatically to match traffic. Which concept does this describe?

A. Fault tolerance B. Vertical scaling
C. Elasticity D. Economies of scale

7. A startup does not want to make a large up-front hardware investment before it knows whether its product will succeed. Which advantage of the AWS Cloud addresses this?

A. Trade fixed expense for variable expense B. Go global in minutes
C. Stop spending money running and maintaining data

centers

D. High availability

8. Which statement correctly describes the difference between high availability and fault tolerance?

A. High availability requires multiple Regions; fault

tolerance requires only one server.

B. Fault tolerance allows continued operation with no

interruption during a failure; high availability minimizes

downtime but may allow a brief interruption.

C. High availability is a property of SaaS; fault tolerance is a

property of IaaS.

D. They are identical terms.

9. Which benefit of cloud computing is made possible because AWS aggregates usage from a very large number of customers?

A. Elasticity B. Agility
C. Economies of scale D. Fault tolerance

10. A company needs a disaster recovery strategy with the LOWEST cost and can tolerate a recovery time of a day. Which strategy should it choose?

A. Multi-site active/active B. Warm standby
C. Pilot light D. Backup and restore

11. In which disaster recovery strategy is the core database continuously replicated to the recovery Region, while application servers remain off until a disaster occurs?

A. Backup and restore B. Pilot light
C. Warm standby D. Multi-site active/active

12. A company states that after an outage it can afford to lose no more than five minutes of data. Which metric does this define?

A. Recovery Time Objective (RTO) B. Recovery Point Objective (RPO)
C. Service level agreement (SLA) D. Mean time between failures (MTBF)

13. Which of the following are advantages of cloud computing as described by AWS? (Choose two.)

A. Go global in minutes B. Eliminate all security responsibilities
C. Increase speed and agility D. Guarantee zero downtime for every application
E. Lock in hardware for three years

14. A company replaces one large EC2 instance with several smaller instances behind a load balancer. What type of scaling is this?

A. Vertical scaling B. Horizontal scaling
C. Scaling up D. Fixed provisioning

15. A research team wants to test a new idea for two days on a large cluster of servers and then discard the cluster. Which cloud benefit does this BEST illustrate?

A. Increase speed and agility B. Economies of scale
C. Fault tolerance D. Hybrid deployment

16. A company wants its application to continue serving users if one Availability Zone becomes unavailable. What is the MOST appropriate design?

A. Run a larger EC2 instance in a single Availability Zone B. Run EC2 instances in multiple Availability Zones behind a load balancer
C. Take daily snapshots of the instance D. Purchase a Reserved Instance

17. Which disaster recovery strategy runs a scaled-down but fully functional copy of the production environment at all times in another Region?

A. Pilot light B. Warm standby
C. Backup and restore D. Cold site

18. A finance team wants its IT staff to stop spending time on hardware refreshes and cooling systems and instead build features for customers. Which statements describe the benefit the team is seeking? (Choose two.)

A. Stop spending money running and maintaining data

centers

B. Focus on what differentiates the business rather than

undifferentiated heavy lifting

C. Go global in minutes D. Use a private cloud
E. Scale vertically

Answers and explanations

1. B. Stop guessing capacity means you no longer have to forecast years ahead; you adjust capacity as needed. Economies of scale (C) is about AWS’s lower prices, and agility (D) is about speed of experimentation, not forecasting.
2. B. This is AWS’s definition of cloud computing. A colocation contract (A) and in-house virtualization (C) are still owned or leased fixed capacity, and installed licenses (D) are not delivered on demand over the internet.
3. B. PaaS, such as AWS Elastic Beanstalk, lets developers upload code while the platform manages servers and scaling. IaaS (A) would require managing the OS; SaaS (C) is a finished application, not a place to deploy your own code.
4. C. Running some resources on-premises and connecting them to resources in the cloud is a hybrid deployment. A private cloud (B) is entirely on-premises, and multi-site (D) is a disaster recovery strategy, not a deployment model.
5. A. Amazon EC2 provides virtual servers where you manage the operating system and above, which is IaaS. Elastic Beanstalk and Lambda are closer to PaaS or serverless, and a hosted CRM application is a finished application (SaaS).
6. C. Elasticity automatically adds and removes resources as demand changes. Vertical scaling (B) changes the size of one resource and does not imply automatic shrinking; fault tolerance (A) is about surviving failures.
7. A. Paying only for what you consume, instead of a large up-front investment, is trading fixed expense for variable expense. The other options are real benefits but do not address up-front cost.
8. B. A fault-tolerant system keeps running with no interruption; a highly available system recovers quickly, possibly with a brief interruption. Neither is tied to a specific service model or number of Regions.
9. C. AWS’s very large aggregate usage lets it achieve lower costs, which it passes on as lower prices. That is economies of scale. Elasticity and agility come from on-demand provisioning, not aggregation.
10. D. Backup and restore is the cheapest strategy and suits workloads that can tolerate hours of recovery time. The other strategies keep resources running in the recovery site and cost more.
11. B. In pilot light, the core (usually the database) is kept live and replicated while other servers are off. Warm standby (C) keeps a small but complete copy running, including servers.
12. B. RPO is the maximum acceptable data loss measured in time. RTO (A) measures how long the outage may last, not how much data is lost.
13. A and C. Both are among the six advantages. The cloud does not eliminate all security responsibilities (B), cannot guarantee zero downtime for every application (D), and avoids hardware lock-in rather than requiring it (E).
14. B. Adding more instances is horizontal scaling (scaling out). Vertical scaling or scaling up (A, C) means making a single instance larger.
15. A. Low-cost, short-lived experiments are the essence of speed and agility. Economies of scale (B) concerns prices, and fault tolerance (C) concerns surviving failures.
16. B. Spreading instances across multiple Availability Zones behind a load balancer keeps the application serving users if one AZ fails. A larger single instance (A) is still a single point of failure; snapshots (C) help recovery but not continuous availability; Reserved Instances (D) are a pricing option.
17. B. Warm standby keeps a smaller, fully functional copy running. Pilot light (A) keeps only the core data live; backup and restore (C) keeps only backups. “Cold site” (D) is a general term, not one of the four AWS strategies.
18. A and B. Moving staff away from running data centers to focus on differentiating work is the advantage “stop spending money running and maintaining data centers”, which removes undifferentiated heavy lifting. Going global (C) and private cloud (D) do not fit the scenario.