Engineering ReferenceEngineering Reference

Cloud Computing

On this page

Cloud computing delivers on-demand compute, storage, networking, and platform services over the internet, billed by usage. It trades upfront capex for opex and engineering effort for managed services.

Overview

The advantage is elasticity and managed services; the trap is sprawl, lock-in, and unbounded bills. Sound architecture, cost tagging, and FinOps practices are as important as the technical design.

Service Models

  • IaaS — VMs, networks, storage (EC2, GCE, Azure VM).
  • PaaS — managed runtimes & databases (App Engine, App Service, Heroku).
  • SaaS — finished applications (Microsoft 365, Salesforce).
  • FaaS / Serverless — event-driven functions (Lambda, Cloud Functions, Azure Functions).
  • CaaS — managed containers (ECS, GKE, AKS, Cloud Run).

Deployment Models

  • Public, private, hybrid, multi-cloud.
  • Edge cloud — compute close to data (CloudFront Functions, Cloudflare Workers).
  • Sovereign / regulated clouds (GovCloud, EU sovereign).

Core Services

  • Compute — VMs, containers, serverless.
  • Storage — object (S3), block (EBS), file (EFS/NFS).
  • Networking — VPC, subnets, load balancers, CDN, VPN, DNS.
  • Databases — RDS, DynamoDB, Cosmos DB, BigQuery.
  • Identity — IAM, roles, policies, federation.
  • Observability — CloudWatch, Stackdriver, Azure Monitor.
  • AI/ML — SageMaker, Vertex AI, Azure ML, Bedrock.

Design Patterns

  • Stateless compute + external state.
  • Auto-scaling groups + load balancers.
  • Multi-AZ for HA; multi-region for DR.
  • Event-driven (SQS, EventBridge, Pub/Sub).
  • Infrastructure as code (Terraform, Pulumi, CloudFormation, Bicep).
  • Least privilege IAM; secret managers; KMS-managed keys.

Providers

  • AWS — broadest service catalog.
  • Microsoft Azure — strong with Microsoft stack & hybrid.
  • Google Cloud — data & AI strengths.
  • Oracle Cloud, IBM Cloud, Alibaba Cloud.
  • Niche / dev: Cloudflare, Fly.io, Vercel, DigitalOcean, Hetzner.
reference page