Getting Started with Cloud

Last reviewed: August 2026

Definition of Cloud

Cloud computing is a service that provides computing resources (servers, storage, networking, and more) on demand over the internet. Used well, it offers advantages in infrastructure build speed, operational efficiency, and cost structure.

From an infrastructure provisioning perspective, the process of purchasing servers, installing them in racks, and configuring networks in your own data center can be replaced with API calls. However, preparatory work such as security review, verifying regulatory requirements, and designing access control is still necessary.

That said, it is difficult to expect the well-known benefits of cost savings or stable operations simply by using the cloud. You need to change your application’s architecture to match how the cloud operates. Changing the architecture and making the most of managed services can streamline operational resources, letting you focus your people and budget on creating business value. However, the more you depend on managed services, the more vendor lock-in increases, so you should also consider an exit strategy.

Core Characteristics of Cloud (Summary)

Here are the five core characteristics of cloud as defined by NIST.

Characteristic	Description	Compared to On-Premises
On-demand self-service	Create resources within minutes via web console/API	Weeks to months to purchase and install → instant
Broad network access	Access from anywhere over the internet using standard protocols	VPN/dedicated line required → not required
Resource pooling	Dynamic allocation via multi-tenant model, region-level location assignment	Fixed allocation per department → shared pool
Rapid elasticity	Automatic scale up/down based on traffic	Pre-provisioned for peak load → real-time adjustment
Measured service	Usage-based pay-as-you-go (per second/hour/GB)	Fixed depreciation cost → variable cost

Why Use Cloud

Here are concrete scenarios, compared with on-premises, showing what practical value the five NIST characteristics create.

Elasticity — Instant response to traffic fluctuations

On-premises, you must purchase additional servers weeks in advance before Black Friday or a new service launch. If your forecast is off, you either run short on servers (outage) or have excess (waste).

In the cloud, autoscaling adds or removes servers within minutes based on traffic. Once the event ends, capacity automatically scales down, reducing cost.

Scenario	On-Premises	Cloud
10x increase in event traffic	Servers must be purchased and installed weeks in advance	Autoscaling responds within minutes
After the event ends	Remaining servers sit idle (depreciation continues)	Automatic scale-down, cost drops immediately
Forecast failure	Server shortage → outage or emergency purchase	Just adjust the limit for instant scaling

Scalability — Global service in minutes

On-premises, launching a service overseas requires months for local data center contracts, server shipping, and network configuration.

In the cloud, you can choose the region you want and deploy the same infrastructure code to go global within minutes.

Scenario	On-Premises	Cloud
Entering the Southeast Asia market	Data center contract + server purchase + networking (3-6 months)	Deploy to the Singapore region (minutes)
Decision to withdraw a service	Equipment disposal, contract termination (penalty fees)	Delete resources (instant, zero cost)

Cost structure shift — From fixed costs to variable costs

On-premises is capital expenditure (CapEx), where depreciation begins the moment you purchase a server. Cloud shifts this to operational expenditure (OpEx), where you pay only for what you use. However, you can only enjoy this benefit if you design your architecture for the cloud. Simply lifting and shifting your on-premises setup onto the cloud as-is can actually increase costs.

Reduced operational burden — Leveraging managed services

On-premises, operating a database requires you to handle OS patching, DB engine updates, backups, replication, and failover all yourself.

Managed cloud services (RDS, Cloud SQL, and so on) let the vendor handle this operational burden instead. DBAs can then focus on schema design and query optimization.

Service Models: IaaS, PaaS, SaaS

Cloud services are divided into three models based on the scope the vendor manages. Compared to your own data center, they work as follows:

IaaS (Infrastructure as a Service) — Similar to renting a building. The building (servers, network, storage) is provided, but you must configure the interior (OS, middleware, applications) yourself.
PaaS (Platform as a Service) — Similar to renting an office. The desks and chairs (runtime, middleware) are already set up, so you can focus solely on the work (application code).
SaaS (Software as a Service) — Similar to staying at a hotel. Everything is ready; you simply use the service as-is.

Service Model	User-Managed Scope	AWS Example	Azure Example	Google Cloud Example
IaaS	OS, middleware, app, data	EC2, EBS, VPC	Virtual Machines, VNet	Compute Engine, VPC
PaaS	App, data	Elastic Beanstalk, RDS	App Service, Azure SQL	App Engine, Cloud SQL
SaaS	Data (settings)	WorkMail, Chime	Microsoft 365, Dynamics 365	Google Workspace

The division of responsibility between the user and the vendor varies by service model. This is covered in detail in Shared Responsibility Model.

Deployment Models: Public, Private, Hybrid

Cloud is divided into three deployment models based on who owns and operates the infrastructure.

Public Cloud

A model where infrastructure owned and operated by a vendor such as AWS, Azure, or Google Cloud is shared by multiple customers over the internet. You can start using it immediately with no upfront investment, and it allows elastic scaling. Most of the content covered in CloudPick is based on public cloud.

Private Cloud

A cloud operated exclusively for a specific organization. It can be built in your own data center, or provided by a vendor as dedicated infrastructure. It is widely used in the finance and public sectors, where security and regulatory requirements are strict.

Hybrid Cloud

A model that connects and uses public cloud and private cloud (or on-premises) together. It is typically operated by placing sensitive data in a private environment while placing elastic workloads in the public cloud.

Deployment Model	AWS	Azure	Google Cloud	OCI
Public	AWS Region	Azure Region	Google Cloud Region	OCI Region
Private/On-Premises Extension	Outposts	Azure Stack, Azure Local	Google Distributed Cloud	Dedicated Region
Hybrid Management	EKS Anywhere, ECS Anywhere	Azure Arc	Anthos	OCI Multicloud

Understanding Multi-Cloud

Using a single cloud vendor is the simplest approach, but in practice, more organizations are adopting a multi-cloud strategy that uses multiple vendors together. According to the CNCF 2024 survey, roughly 60% of enterprises use two or more clouds.

The motivations and challenges of adopting multi-cloud are covered in detail in Understanding Multi-Cloud.

CloudPick aims to help you make sound decisions in a multi-cloud environment, centered around the major global cloud vendors.

Common Mistakes

“Moving to the cloud automatically reduces costs” — Lifting and shifting your on-premises structure as-is can actually increase costs. A shift to cloud-native architecture is required.
“The cloud vendor manages everything for you” — The user’s scope of responsibility differs depending on the service model (IaaS/PaaS/SaaS). Data and access control are always the user’s responsibility.
“The free tier is enough to run on” — The free tier is meant for learning and PoCs. Exceeding the free limits can result in unexpected charges, so set up budget alerts.

Checklist

Have you checked the free tier scope and limits of the vendor you plan to use?
Have you enabled MFA (multi-factor authentication) on the root/admin account?
Have you set up budget alerts to prevent unexpected charges?

References

Standards and Frameworks

NIST SP 800-145 — The NIST Definition of Cloud Computing — Official definition of cloud computing
ISO/IEC 17788 — Cloud computing: Overview and vocabulary — Cloud terminology standard
ISO/IEC 22123 — Cloud computing: Concepts and terminology — Latest standard including multi-cloud