The Different Types of Cloud Computing

Key Takeaways​

Cloud computing isn't one-size-fits-all; it operates on a spectrum of responsibility split between you and AWS. Think of it like building with LEGO bricks: IaaS gives you the raw, unconfigured blocks (EC2); PaaS gives you a pre-assembled frame where you just drop in your code (Elastic Beanstalk); and SaaS gives you a fully finished, ready-to-use toy (Amazon Rekognition or Zoom).

Financially, AWS boils cloud billing down to three core drivers: Compute (pay per second/hour), Storage (pay per GB), and Data Transfer Out (inbound is always free).

[ Traditional On-Prem ]  ---> You manage everything (Hardware, OS, App)
[ IaaS ]                 ---> AWS manages Hardware/Virtualization; You manage OS & App
[ PaaS ]                 ---> AWS manages Hardware & OS; You manage Code & Data
[ SaaS ]                 ---> AWS manages everything; You just consume the service

Main Discussion​

The Shared Responsibility Gradient: Service Model Breakdown​

Understanding cloud service models means understanding who manages what in the technology stack.

• Infrastructure as a Service (IaaS): Provides raw networking, virtual machines, and storage. Gives maximum flexibility and architectural control.
  • AWS Example: Amazon EC2
  • Industry Examples: Google Cloud Compute Engine, Microsoft Azure VMs, Linode, DigitalOcean
• Platform as a Service (PaaS): Abstracts away OS patching, hardware provisioning, and runtime management so developers can focus solely on application deployment and code.
  • AWS Example: AWS Elastic Beanstalk
  • Industry Examples: Heroku, Google App Engine
• Software as a Service (SaaS): Fully managed, end-user applications delivered over the internet. You configure settings and consume capabilities without touching underlying code or infrastructure.
  • AWS Example: Amazon Rekognition (pre-trained AI vision API)
  • Industry Examples: Gmail, Zoom, Dropbox

The Operational Responsibility Equation​

We can express the management burden on your internal IT operations team as a function of the cloud model chosen:

$\text{User Management Burden} = \text{Total Stack} - \text{Provider Managed Layer}$

$\text{IaaS Burden} = \text{Application} + \text{Data} + \text{Runtime} + \text{Middleware} + \text{OS}$

$\text{PaaS Burden} = \text{Application} + \text{Data}$

$\text{SaaS Burden} = \text{Zero Infrastructure / Zero Code (Configuration Only)}$

Deciphering the AWS Financial Engine: The 3 Pricing Pillars​

AWS eliminates upfront hardware costs by driving all consumption through a utility-style pay-as-you-go model.

                      +----------------------------------+
                      |   AWS 3 PRICING FUNDAMENTALS     |
                      +----------------------------------+
                                       |
         +-----------------------------+-----------------------------+
         |                             |                             |
         v                             v                             v
  [ 1. COMPUTE ]                 [ 2. STORAGE ]              [ 3. NETWORKING ]
Charged per second/hour        Charged per GB stored        Pay for Data OUT only
(EC2, Lambda, Bedrock)           (S3, EBS, DynamoDB)          (Data IN is FREE!)

$\text{Total Monthly AWS Cost} = f(\text{Compute Duration}) + f(\text{GB Stored}) + f(\text{GB Data Transferred OUT})$

1. Compute: Billed based on execution duration and instance size (e.g., clock hours or seconds an EC2 instance or AI inference task runs).
2. Storage: Billed based on the total volume of data (in Gigabytes/Terabytes) stored persistently in services like Amazon S3 or EBS.
3. Networking (Data Transfer):
   • Inbound Data Transfer (Data IN): 100% Free across all AWS services.
   • Outbound Data Transfer (Data OUT): Aggregated across services and charged on a per-GB basis as data leaves the AWS Cloud.

Exam Guide​

Exam Tips​

• Identify Service Models for AI Services: Pay close attention to how AWS AI services are categorized on the exam:
  • Amazon EC2 / EBS: IaaS (you manage the virtual machine and OS).
  • AWS Elastic Beanstalk / SageMaker Managed Endpoints: PaaS (managed runtime environment for applications/models).
  • Amazon Rekognition / Amazon Polly / Amazon Transcribe: SaaS (turnkey, pre-trained AI APIs where you pass data and get instant results without managing models or servers).
• Data Transfer In vs. Out: A classic trap question involves egress fees. Remember: Data coming into AWS is always free. You only pay when data leaves the AWS region or cloud environment to the internet.

Practice Test​

Question 1​

A developer wants to integrate image classification capabilities into a web application. They prefer to send images to an existing pre-trained AWS API and receive JSON label predictions without managing training jobs, virtual machines, or operating system patches. Which cloud service model and AWS service best fit this requirement?

• A. IaaS using Amazon EC2
• B. PaaS using AWS Elastic Beanstalk
• C. SaaS using Amazon Rekognition
• D. Hybrid Cloud using AWS Outposts

Question 2​

A cloud practitioner is reviewing a monthly AWS invoice and notices a line item for network charges. Which network traffic pattern incur a fee under standard AWS billing rules?

• A. Uploading 500 GB of training datasets from an on-premises workstation into an Amazon S3 bucket
• B. Inbound data transfer from the internet into an Amazon EC2 instance
• C. Transferring data out from an Amazon S3 bucket to an external internet destination
• D. Data ingress coming through a public endpoint into an AI processing pipeline