Cloud Service Model · 1 of 6

IaaS — Infrastructure as a Service

Rent the raw building blocks — virtual machines, networks, storage, load balancers — and assemble whatever you want on top. The most flexible, lowest-level cloud model.

VMsNetworksStorageYou manage the OSMaximum control
← Back to Cloud
Quick Facts

At a Glance

Basic Concepts

  • You rent the hardware (virtually) and own everything from the OS up.
  • Pay per second / hour for compute; per GB-month for storage; per GB for egress.
  • Elastic: spin up 100 servers in 5 minutes; tear them down in 1.
  • Building blocks: Compute (VM), Storage (disk + object), Networking (VPC, LB).
  • Closest to "data center as code" — you control firewalls, subnets, IAM, encryption.
Building Blocks

What IaaS Gives You

CategoryExamples
Virtual machinesAWS EC2, Azure VMs, GCP Compute Engine, OCI Compute
Block storageEBS, Azure Managed Disks, Persistent Disk
Object storageS3, Azure Blob, GCS, R2 (Cloudflare)
File storageEFS, Azure Files, Filestore
NetworkingVPC, subnets, security groups, route tables, NAT, peering
Load balancersALB / NLB, Azure LB, GCLB
DNSRoute 53, Azure DNS, Cloud DNS
CDNCloudFront, Azure Front Door, Cloud CDN
IdentityIAM (AWS), Entra ID, Cloud IAM
Bare metalEC2 metal, Azure dedicated, Bare Metal Solution
Mechanics

Working with IaaS

Provisioning & Infrastructure as Code

You almost never click in the console for production. Define infra in code:

  • Terraform / OpenTofu — cloud-agnostic, declarative.
  • AWS CloudFormation / CDK — AWS-native.
  • Azure Bicep / ARM — Azure-native.
  • Pulumi — IaC in real languages (TS, Python, Go).
  • Ansible — for what happens inside the VM.
VM Sizing & Families

Each cloud has dozens of VM "families" optimized for different workloads:

  • General purpose (M-series, D-series) — balanced CPU/RAM.
  • Compute optimized (C-series) — CPU-heavy.
  • Memory optimized (R-series, X-series) — in-memory DBs, analytics.
  • Storage optimized (I-series) — local NVMe SSDs.
  • GPU / accelerator (P / G / N-series) — AI training & inference.
  • ARM (Graviton) — cheaper, power-efficient.
Pricing Models & Cost Levers
  • On-demand — pay per second, no commitment, highest rate.
  • Reserved / Savings Plans — 1- or 3-year commit for 30–70% off.
  • Spot / Preemptible — 70–90% off, can be reclaimed in minutes; great for batch.
  • Egress is the silent killer — moving data out often dwarfs compute cost.
  • Right-sizing + autoscaling are typically the biggest wins.
Operational Responsibilities
You handleCloud handles
OS patching, hardeningHypervisor, hardware
App deployment, runtimeNetwork fabric, power
Backups (mostly)Storage durability (11 nines)
Firewall rules, IAMPhysical security, DDoS-edge
Monitoring & loggingRegion/AZ infrastructure
Trade-offs

Strengths & Weaknesses

Strengths
  • Maximum control & flexibility — runs anything.
  • Predictable performance (no shared multi-tenant runtime).
  • Smooth lift-and-shift from on-prem.
  • Compliance-friendly with dedicated tenancy.
Weaknesses
  • You own all the ops — patching, monitoring, scaling.
  • Slowest path from idea to production.
  • Cost spirals fast without disciplined right-sizing.
  • Easy to drift from cloud-native best practices.
When to Use

Sweet Spots

Lift-and-Shift

Move existing workloads with minimal rewrite.

Custom / Legacy Stacks

Apps that need specific OS / kernel / driver versions.

Compliance-Heavy

Dedicated tenancy, full audit control.

HPC / AI Training

GPU clusters, MPI, low-latency networking.

Continue

Other Service Models

PaaSSaaSFaaS / ServerlessDBaaSContainers as a Service ↑ Back to Cloud