☁️ Cloud Service Provider Comparison
Every deployment decision starts with a question: where does this app actually run? This lesson maps the cloud landscape — the service models, the big three providers, and the lean developer-friendly platforms — so you can pick infrastructure that fits your project instead of following hype.
🎯 Learning Objectives
By the end of this lesson, you will be able to:
- Explain the five cloud service models (IaaS, PaaS, SaaS, FaaS, CaaS) and who manages what in each
- Compare AWS, Azure, and Google Cloud by strengths, flagship services, and pricing style
- Map an equivalent service across providers (compute, storage, database, containers, serverless)
- Apply a selection framework weighing technical, business, and strategic factors
- Describe multi-cloud and hybrid strategies and the trade-offs each brings
Estimated Time: 30–40 minutes • Difficulty: Intermediate
Hands-on: Draft a provider recommendation for a real e-commerce migration scenario.
In This Lesson
What Cloud Computing Buys You
Cloud computing lets you rent computing resources on demand instead of buying and babysitting physical hardware. You provision a server in seconds, pay only for what you use, and hand off the electricity, cooling, security patching, and hardware replacement to someone else.
💡 The electricity analogy: You don't run a generator in your basement — you connect to the grid, flip a switch, and pay for the kilowatt-hours you consume. The cloud is the grid for compute. You tap in, use what you need, and the utility handles the messy infrastructure underneath.
That shift changes what you optimize for. On-premises, the expensive mistake is under-buying hardware and running out of capacity. In the cloud, the expensive mistake is leaving resources running that you no longer use. The workflow inverts from "provision for peak, forever" to "scale to demand, continuously":
✅ What you gain
- Cost model shift — capital expense (buy servers) becomes operating expense (rent by the hour)
- Elastic scale — grow and shrink capacity in minutes, not procurement cycles
- Global reach — deploy close to users in dozens of regions worldwide
- Managed reliability — redundant power, networking, and hardware you never touch
- Access to advanced services — managed databases, ML, and analytics without building them yourself
The Five Service Models
Cloud services come in layers. The higher up the stack you go, the more the provider manages and the less you control. The classic mental model is the "pizza as a service" ladder — how much you make yourself versus how much is handled for you.
📖 The five models
IaaS (Infrastructure as a Service): raw virtual machines, storage, and networking. You manage the OS and everything above. Like renting an empty apartment. — AWS EC2, Azure VMs, Compute Engine.
PaaS (Platform as a Service): you push code; the platform runs it. Like a furnished apartment. — Elastic Beanstalk, App Engine, Azure App Service, Heroku.
SaaS (Software as a Service): ready-to-use software over the web. Like a hotel room. — Google Workspace, Salesforce, Microsoft 365.
FaaS / Serverless (Function as a Service): run individual functions on events, pay per invocation. Like food delivery — you pay only when you order. — AWS Lambda, Cloud Functions, Azure Functions.
CaaS (Container as a Service): a managed home for your containers. Like a co-working space. — AWS ECS/EKS, Google Kubernetes Engine, Azure Kubernetes Service.
⚠️ "Serverless" still has servers. The name means you don't manage them — the provider provisions, scales, and patches the machines invisibly. You're billed for execution time, not idle capacity.
The Big Three Providers
Three providers dominate the market. Each can host almost any application, but each has a distinct personality worth knowing before you commit.
Amazon Web Services (AWS)
The first mover and still the broadest. AWS offers 200+ services and the largest global footprint. It's the safe default for "we might need anything later," at the cost of famously intricate pricing.
- Flagships: EC2 (compute), S3 (object storage), RDS (managed SQL), Lambda (serverless), ECS/EKS (containers)
- Pricing: pay-as-you-go with on-demand, Reserved Instances, and Spot Instances — powerful but hard to predict without tooling
- Best when: you want maximum service breadth and a huge hiring pool
Microsoft Azure
The enterprise and hybrid-cloud favorite. If your organization already runs Windows Server, Active Directory, and Microsoft 365, Azure integrates seamlessly and often comes bundled into existing agreements.
- Flagships: Azure Virtual Machines, App Service (PaaS), Azure SQL Database, Azure Functions, AKS
- Pricing: consumption-based, with meaningful discounts for existing Microsoft/enterprise customers
- Best when: you're a Microsoft shop or need strong hybrid (on-prem + cloud) support
Google Cloud Platform (GCP)
Built on Google's own network, GCP leads in data analytics, machine learning, and Kubernetes (which Google created). Its pricing tends to be simpler, with automatic sustained-use discounts.
- Flagships: Compute Engine, GKE (managed Kubernetes), BigQuery (analytics), Cloud Functions, Vertex AI
- Pricing: pay-as-you-go with per-second billing and automatic sustained-use discounts
- Best when: your workload is data/ML-heavy or Kubernetes-native
💡 Beyond the big three
Not every project needs a hyperscaler. DigitalOcean, Render, Fly.io, Railway, and Heroku trade breadth for simplicity and predictable pricing — often the right call for solo developers, startups, and MVPs. We deploy to Heroku two lessons from now.
Side-by-Side Comparison
The single most useful skill here is translation: recognizing that "S3", "Blob Storage", and "Cloud Storage" are the same idea wearing three brand names. Learn one provider deeply and this table lets you read the others.
| Service Type | AWS | Azure | GCP |
|---|---|---|---|
| Compute (VMs) | EC2 | Virtual Machines | Compute Engine |
| Object storage | S3 | Blob Storage | Cloud Storage |
| Managed SQL | RDS | SQL Database | Cloud SQL |
| NoSQL | DynamoDB | Cosmos DB | Firestore |
| Containers (K8s) | EKS | AKS | GKE |
| Serverless | Lambda | Functions | Cloud Functions |
| PaaS web hosting | Elastic Beanstalk | App Service | App Engine |
| CDN | CloudFront | Front Door / CDN | Cloud CDN |
| ML platform | SageMaker | Azure ML | Vertex AI |
Rough pricing snapshot
Prices shift constantly and vary by region and commitment. Treat these as order-of-magnitude anchors, not quotes — always confirm on the provider's current calculator.
| Item | AWS | Azure | GCP |
|---|---|---|---|
| VM (2 vCPU, 8 GB), on-demand | ~$60–70/mo (t3.large) | ~$70/mo (D2s v3) | ~$50–58/mo (e2-standard-2) |
| Object storage, 1 TB/mo | ~$23 (S3 Standard) | ~$20 (Blob Hot) | ~$20 (Standard) |
| Serverless, 1M invocations | ~$0.20 (Lambda) | ~$0.20 (Functions) | ~$0.40 (Cloud Functions) |
| Egress, 100 GB out | ~$9.00 | ~$8.50 | ~$12.00 |
⚠️ Egress is the hidden bill. Getting data into the cloud is usually free. Getting it out — and moving it between regions or providers — is where surprise costs live. Design your architecture to keep chatty traffic inside one region.
Choosing a Provider
There is no universally "best" provider — only the best fit for a given set of constraints. Weigh three categories of factors:
When you're stuck, this quick heuristic points you at a starting shortlist:
| If you need… | Start with… | Because… |
|---|---|---|
| Broadest service catalog | AWS | Largest, most mature portfolio and hiring pool |
| Microsoft / enterprise integration | Azure | Deep ties to AD, Windows, and Microsoft 365 |
| Data analytics & ML | GCP | BigQuery and Vertex AI lead the field |
| Strong hybrid (on-prem + cloud) | Azure | Azure Arc / Stack designed for it |
| Simplicity & predictable pricing | DigitalOcean / Render / Heroku | Flat pricing, gentle learning curve |
| Global edge presence | Fly.io / AWS | Deploy close to users worldwide |
⚠️ Watch for vendor lock-in
The more proprietary managed services you adopt (DynamoDB, Cosmos DB, BigQuery), the faster you build but the harder you'll find leaving. Portable choices — containers, PostgreSQL, open standards — cost a little more effort now and buy you an exit later. Decide deliberately which lock-in you're willing to accept.
Multi-Cloud & Hybrid
You're not limited to one provider. Two broader strategies come up constantly on real teams.
Multi-cloud
Using more than one provider at once — say, AWS for compute, GCP for analytics, Azure for enterprise apps.
Pros: best-of-breed services, reduced lock-in, resilience against a single provider's outage. Cons: real operational complexity, cross-cloud egress costs, and the need for skills across multiple platforms.
Hybrid cloud
Combining public cloud with on-premises or private infrastructure — common where data sovereignty, compliance, or legacy systems keep some workloads in your own datacenter.
Pros: keep sensitive data in-house, integrate legacy systems, burst to public cloud at peak. Cons: complex networking, harder to manage consistently, higher baseline cost.
✅ The tools that make it manageable
- Terraform — Infrastructure as Code that works across every provider
- Kubernetes — a portable container runtime that abstracts the underlying cloud
- Datadog / New Relic / Grafana — unified monitoring across environments
- GitHub Actions / GitLab CI — provider-agnostic deployment pipelines
Hands-on Exercise
🏋️ Provider Recommendation for a Migration
Scenario: You're the tech lead at a growing e-commerce company moving off on-prem hardware. The application is:
- React frontend
- Node.js API backend
- PostgreSQL database + Redis cache
- Product-image storage
- An ML-based recommendation engine
Traffic could grow 5–10× in the next year. Produce a one-page recommendation.
Your tasks
- List the technical requirement for each component (hosting, scaling, storage, ML).
- Map each requirement to a specific service on two providers.
- Note one pro and one con per provider for this workload.
- Make a recommendation and justify it in two sentences.
💡 Hint
Start from the component that's hardest to move. The ML recommendation engine and the managed PostgreSQL are the two that most differentiate providers — the React frontend can go on a CDN almost anywhere. Let the specialized parts drive the decision.
✅ Example answer (AWS vs GCP)
| Component | AWS | GCP |
|---|---|---|
| Frontend | S3 + CloudFront | Cloud Storage + Cloud CDN |
| API backend | ECS Fargate | Cloud Run / GKE |
| Database | RDS for PostgreSQL | Cloud SQL for PostgreSQL |
| Cache | ElastiCache for Redis | Memorystore for Redis |
| Image storage | S3 | Cloud Storage |
| Recommendations | SageMaker | Vertex AI |
Recommendation: Go with GCP if the recommendation engine is the strategic differentiator — Vertex AI and BigQuery give the data team the strongest tooling, and Cloud Run keeps the API simple. Choose AWS instead if the team already has AWS skills, since breadth and a larger hiring pool reduce delivery risk during a fast-growth year.
🎯 Quick Quiz
Question 1: In which service model does the provider manage the most — leaving you responsible only for using ready-made software?
Question 2: Which provider is most often chosen specifically for its data analytics and machine learning strengths?
Question 3: What is the main downside of adopting many proprietary managed services (e.g. DynamoDB, BigQuery)?
Summary & Quiz
🎉 Key Takeaways
- The cloud trades capital expense and hardware toil for elastic, pay-as-you-go resources.
- Service models (IaaS → CaaS → PaaS → FaaS → SaaS) shift responsibility from you to the provider as you climb.
- AWS = breadth, Azure = enterprise/hybrid, GCP = data & ML; leaner platforms trade breadth for simplicity.
- Equivalent services exist across providers — learn one deeply, then translate.
- Choose by weighing technical, business, and strategic factors; watch egress costs and lock-in.
📚 Further Reading
- AWS — Getting Started
- Microsoft — Cloud Adoption Framework
- Google Cloud — Documentation Overview
- The Twelve-Factor App
🚀 What's Next?
Now that you can pick a provider, we'll go deep on the market leader. Next up: AWS Deployment Strategies — EC2, containers, serverless, and the CI/CD pipelines that ship to them.
🎉 Great work!
You can now read the cloud landscape and defend a provider choice. Let's put an app on one.