⏳
Loading cheatsheet...
Cloud service models, migration patterns, provider selection, shared responsibility, and practical cloud decision frameworks.
| Model | Full Name | You Manage | Provider Manages | Examples |
|---|---|---|---|---|
| IaaS | Infrastructure as a Service | OS, runtime, apps, data | Hardware, networking, virtualization | EC2, VMs, Compute Engine |
| PaaS | Platform as a Service | Apps, data | OS, runtime, middleware, hardware | App Engine, Heroku, Azure App Service |
| SaaS | Software as a Service | Nothing (just configure) | Everything | Gmail, Salesforce, Office 365 |
| FaaS | Function as a Service | Code only | Everything else (serverless) | Lambda, Functions, Cloud Functions |
| CaaS | Container as a Service | Containers & images | Orchestration, infra | EKS, AKS, GKE |
| Model | Description | Best For |
|---|---|---|
| Public Cloud | Shared infrastructure, multi-tenant | Startups, variable workloads, cost optimization |
| Private Cloud | Dedicated infrastructure, single-tenant | Regulated industries, sensitive data, legacy apps |
| Hybrid Cloud | Public + private with connectivity | Enterprises needing both flexibility and control |
| Multi-Cloud | Using 2+ cloud providers simultaneously | Vendor lock-in avoidance, best-of-breed services |
| Layer | IaaS | PaaS | SaaS |
|---|---|---|---|
| Physical Infrastructure | Provider | Provider | Provider |
| Networking (hardware) | Provider | Provider | Provider |
| Virtualization | Provider | Provider | Provider |
| Operating System | Customer | Provider | Provider |
| Runtime / Middleware | Customer | Provider | Provider |
| Application | Customer | Customer | Provider |
| Data | Customer | Customer | Customer |
| Aspect | CapEx (Capital) | OpEx (Operational) |
|---|---|---|
| Definition | Upfront investment in assets | Pay-as-you-go recurring costs |
| Cloud model | Reserved instances (1-3 yr) | On-demand, pay per use |
| Cash flow | Large upfront, depreciated over time | Predictable monthly costs |
| Flexibility | Hard to change | Easily scale up/down |
| Tax | Depreciation benefits | Full deduction in same year |
| Risk | Over-provisioning waste | Unpredictable spikes |
| Strategy | Description | Effort | Risk | Example |
|---|---|---|---|---|
| Rehost (Lift & Shift) | Move as-is to cloud VMs | Low | Low | Migrating legacy app to EC2 |
| Replatform (Lift & Reshape) | Minor optimizations, same architecture | Medium | Low | Moving from self-hosted DB to RDS |
| Refactor / Re-architect | Redesign for cloud-native | High | Medium | Monolith to microservices on Lambda |
| Repurchase (Replace) | Switch to SaaS solution | Low | Medium | Replace custom CRM with Salesforce |
| Retire | Decommission unused systems | None | None | Shut down old test environments |
| Retain (Keep On-Prem) | Keep in data center | None | Low | Apps with strict data residency laws |
| Service | Category | Description |
|---|---|---|
| EC2 | Compute | Virtual servers — widest instance type selection (300+ types) |
| S3 | Storage | Object storage — 11 nines durability, lifecycle policies, 7 storage classes |
| RDS | Database | Managed relational DB (MySQL, PostgreSQL, MariaDB, Oracle, SQL Server) |
| Lambda | Serverless | Event-driven compute — up to 15 min, 1M free requests/mo |
| VPC | Networking | Virtual private cloud — subnets, security groups, NACLs, NAT gateways |
| IAM | Security | Identity & access management — policies, roles, MFA, federation |
| CloudFront | CDN | Global content delivery — 600+ edge locations, Lambda@Edge |
| Route 53 | DNS | Managed DNS — latency-based routing, health checks |
| Service | Category | Description |
|---|---|---|
| ECS / EKS | Containers | ECS (proprietary), EKS (managed Kubernetes) |
| ElastiCache | Cache | Managed Redis or Memcached |
| Aurora | Database | MySQL/PostgreSQL-compatible, 5x faster, auto-scaling |
| DynamoDB | NoSQL | Single-digit ms latency, auto-scaling, global tables |
| SageMaker | ML | Full ML lifecycle — build, train, deploy models |
| Bedrock | GenAI | Access Claude, Llama, Titan models via API |
| Rekognition | AI | Image & video analysis — face detection, labels, text |
| Redshift | Analytics | Petabyte-scale data warehouse — columnar storage |
| Model | Discount | Commitment | Best For |
|---|---|---|---|
| On-Demand | None | None (pay per second) | Short-term, unpredictable workloads |
| Reserved Instances | Up to 72% off | 1 or 3 year term | Steady-state production workloads |
| Spot Instances | Up to 90% off | None (can be interrupted) | Batch jobs, CI/CD, fault-tolerant workloads |
| Savings Plans | Up to 72% off | 1 or 3 yr $/hr commitment | Flexible compute usage across EC2/Lambda/Fargate |
| Service | Category | Description |
|---|---|---|
| Virtual Machines | Compute | Full VMs — wide OS selection, hybrid benefit |
| Blob Storage | Storage | Object storage — hot, cool, cold, archive tiers |
| Azure SQL | Database | Managed SQL Server — auto-tuning, geo-replication |
| Azure Functions | Serverless | Event-driven — 1M free executions/mo |
| AKS | Containers | Managed Kubernetes — integrates with Azure AD, Monitor |
| App Service | PaaS | Fully managed web apps — .NET, Java, Node, Python |
| CosmosDB | NoSQL | Multi-model, globally distributed, 5 consistency levels |
| Azure Cache for Redis | Cache | Managed Redis — Enterprise tier with persistence |
| Service | Category | Description |
|---|---|---|
| Entra ID (Azure AD) | Identity | SSO, MFA, conditional access — integrates with Office 365 |
| Intune | MDM | Device management — BYOD policies, compliance |
| Defender | Security | Cloud security posture, threat protection suite |
| Azure OpenAI | AI | GPT-4, GPT-4o, DALL-E via Azure (enterprise-grade) |
| Azure ML | ML | AutoML, designer (drag-drop), managed endpoints |
| Cognitive Services | AI | Vision, Speech, Language, Translator APIs |
| Azure DevOps | DevOps | Boards, Repos, Pipelines, Test Plans, Artifacts |
| Power Platform | Low-Code | Power BI, Power Apps, Power Automate |
| Service | Category | Description |
|---|---|---|
| Compute Engine | Compute | VMs — custom machine types, per-second billing |
| Cloud Storage | Storage | Object storage — Standard, Nearline, Coldline, Archive |
| Cloud SQL | Database | Managed MySQL, PostgreSQL, SQL Server |
| Cloud Functions | Serverless | Event-driven — Node.js, Python, Go, Java |
| Cloud Run | Containers | Best serverless containers — scale to zero, any language |
| GKE | Containers | Best-in-class managed Kubernetes — 4x leader in Forrester Wave |
| Firestore | NoSQL | Serverless document DB — real-time sync, offline mode |
| BigQuery | Analytics | Best data warehouse — separate storage & compute, $5/TB queried |
| Service | Category | Description |
|---|---|---|
| Vertex AI | ML Platform | Full ML lifecycle — AutoML, custom training, model registry |
| Gemini (API) | GenAI | Gemini Pro/Ultra — multimodal, best pricing for inference |
| TPUs | Hardware | Custom AI accelerators — 2-8x faster than GPUs for training |
| AutoML | ML | No-code ML — vision, NLP, tabular, translation models |
| Dataflow | Streaming | Managed Apache Beam — batch & streaming ETL |
| Pub/Sub | Messaging | Global event streaming — millions of messages/sec |
| Spanner | Database | Globally consistent relational DB — 99.999% availability |
| Cloud CDN | CDN | Global CDN — 150+ edge locations, Google backbone network |
| Resource | Free Allowance | Notes |
|---|---|---|
| e2-micro VM | 1 instance (US regions) | 30 GB standard persistent disk |
| Cloud Storage | 5 GB Standard | 1 GB/day Class A, 5 GB/day Class B operations |
| BigQuery | 10 GB storage | 1 TB query/month free |
| Cloud Functions | 2M invocations/month | 400K GB-sec, 200K GB-sec network egress |
| Cloud Run | 2M requests/month | 360K vCPU-sec, 180K GB-hr memory |
| Firestore | 1 GB storage | 50K reads, 20K writes, 20K deletes/day |
| Category | AWS | Azure | GCP |
|---|---|---|---|
| VMs | EC2 (300+ types) | Virtual Machines (200+ types) | Compute Engine (custom types) |
| VM pricing | t3.micro ~$8/mo | B1s ~$7/mo | e2-micro ~$6/mo |
| Serverless | Lambda | Functions | Cloud Functions + Cloud Run |
| Serverless timeout | 15 min | 230 sec (consumption) | 60 min (Cloud Run) |
| Object storage | S3 ($0.023/GB) | Blob ($0.018/GB) | Cloud Storage ($0.020/GB) |
| Storage classes | 7 (S3, Glacier) | 4 (Hot, Cool, Cold, Archive) | 4 (Standard, Nearline, Coldline, Archive) |
| Block storage | EBS ($0.08/GB) | Managed Disk ($0.096/GB) | Persistent Disk ($0.08/GB) |
| Category | AWS | Azure | GCP |
|---|---|---|---|
| Managed SQL | RDS | Azure SQL | Cloud SQL |
| Best SQL | Aurora (5x faster) | Azure SQL (auto-tuning) | Cloud SQL (PostgreSQL) |
| NoSQL | DynamoDB | CosmosDB | Firestore / Spanner |
| Cache | ElastiCache | Azure Cache for Redis | Memorystore |
| Kubernetes | EKS ($0.10/hr/cluster) | AKS (free control plane) | GKE (free Autopilot tier) |
| Container serverless | Fargate | Container Apps | Cloud Run (best) |
| Category | AWS | Azure | GCP |
|---|---|---|---|
| ML Platform | SageMaker | Azure ML | Vertex AI (best) |
| GenAI Models | Bedrock (Claude, Titan) | Azure OpenAI (GPT-4) | Gemini (multimodal) |
| Data Warehouse | Redshift ($0.25/TB) | Synapse ($0.077/TB) | BigQuery ($5/TB queried) |
| Data Lake | S3 + Glue + Athena | ADLS + Synapse | GCS + Dataflow |
| Streaming | Kinesis | Event Hubs | Pub/Sub (best) |
| ETL/Orchestration | Step Functions / Glue | Data Factory | Cloud Workflows / Composer |
| Category | AWS | Azure | GCP |
|---|---|---|---|
| CDN | CloudFront (600+ PoPs) | Azure CDN (100+ PoPs) | Cloud CDN (150+ PoPs) |
| DNS | Route 53 | Azure DNS | Cloud DNS |
| Load Balancer | ALB / NLB / GWLB | App GW / LB / Front Door | Cloud LB (global, anycast) |
| Private Connect | PrivateLink / VPC Endpoints | Private Link | Private Service Connect |
| Dedicated Connect | Direct Connect | ExpressRoute | Cloud Interconnect |
| Global Network | 70+ PoPs | 60+ regions | 35+ regions, Google backbone |
| Strategy | Savings | How |
|---|---|---|
| Rightsizing | 20-40% | Match instance types to actual CPU/RAM usage |
| Reserved Instances | 40-72% | Commit to 1-3 years for predictable workloads |
| Spot/Preemptible | 60-90% | Use for batch, CI/CD, fault-tolerant workloads |
| Savings Plans | 40-72% | Flexible $/hr commitment across services |
| Auto-scaling | 30-50% | Scale down during nights/weekends automatically |
| Storage tiering | 40-80% | Move cold data to cheaper storage classes |
| Data transfer | Significant | Minimize cross-AZ, cross-region, and egress traffic |
| Cost Category | AWS | Azure | GCP |
|---|---|---|---|
| Data egress | $0.09/GB (internet) | $0.087/GB | $0.105/GB |
| Cross-AZ transfer | $0.01-0.02/GB | $0.01/GB | $0.01/GB |
| Cross-region transfer | $0.02/GB+ | $0.02/GB | $0.08/GB (higher!) |
| API calls (storage) | $0.005 per 1K | $0.004 per 10K | $0.005 per 1K |
| NAT Gateway | $0.045/hr + $0.045/GB | Not needed (outbound free) | Cloud NAT $0.045/hr |
| Public IP idle | $0.005/hr | Free | $0.004/hr |
| Load Balancer | $0.0225/hr (ALB) | $0.018/hr (Basic) | $0.025/hr |
| Resource | AWS Free Tier | Azure Free Tier | GCP Always Free |
|---|---|---|---|
| Compute | 750 hrs t2.micro (12 mo) | 750 hrs B1s (12 mo) | e2-micro (forever) |
| Storage | 5 GB S3 (12 mo) | 5 GB Blob (12 mo) | 5 GB Cloud Storage (forever) |
| Database | 750 hrs db.t2.micro (12 mo) | 250 hrs SQL DB (12 mo) | Cloud SQL (30 day trial) |
| Serverless | 1M Lambda req/mo (12 mo) | 1M Functions req/mo (forever) | 2M req/mo (forever) |
| Network | 100 GB out (12 mo) | 100 GB out (12 mo) | 1 GB out (forever) |
| Analytics | Limited | Limited | 10 GB BigQuery storage (forever) |
| Program | Credits | Requirements |
|---|---|---|
| AWS Activate | $1K-$100K credits | Portfolio company of approved VC/accelerator |
| Azure for Startups | $25K (Founders Hub) | Join Founders Hub — no VC required |
| Google for Startups | $2K-$200K credits | Apply via Google for Startups program |
| Stage | Monthly Budget | Recommended Stack |
|---|---|---|
| Pre-seed / MVP | $0-$100 | Vercel/Netlify (frontend) + Supabase/Firebase (backend) + GCP e2-micro |
| Seed ($1M-$5M) | $100-$5K | AWS/GCP — Fargate/Cloud Run + RDS/Cloud SQL + S3/GCS |
| Series A ($5M-$20M) | $5K-$20K | Multi-service: EKS/GKE + Aurora/CosmosDB + Redis + CDN |
| Series B+ ($20M+) | $20K-$100K+ | Multi-region, dedicated support, reserved instances |
| Standard | AWS | Azure | GCP |
|---|---|---|---|
| HIPAA | Yes | Yes | Yes |
| SOC 1/2/3 | Yes | Yes | Yes |
| PCI-DSS | Yes | Yes | Yes |
| GDPR | Yes | Yes | Yes |
| FedRAMP (High) | Yes | Yes | Moderate only |
| ISO 27001 | Yes | Yes | Yes |
| IRAP (Australia) | Yes | Yes | Yes |
| # of certifications | 143 | 100+ | 125 |
| Feature | AWS | Azure | GCP |
|---|---|---|---|
| Basic Support | Free (forums) | Free (billing + subscriptions) | Free (community) |
| Developer Support | $29/mo | Not available | $29/mo |
| Business Support | $100/mo (10% spend) | $100/mo (min) | $250/mo |
| Enterprise Support | $15K/mo (min) | $1K/mo (min) | Custom pricing |
| SLA (compute) | 99.99% | 99.99% | 99.99% |
| SLA (storage) | 99.999999999% (S3) | 99.999999999% (Blob) | 99.999999999% |
| Dedicated engineer | TAM (Enterprise) | CSAM (Unified) | Customer Engineer |
| Accelerator | Provider | Best For | Price (on-demand) |
|---|---|---|---|
| A10G (24 GB) | AWS / Azure / GCP | Inference, light training | ~$1.00-1.50/hr |
| A100 (40/80 GB) | AWS / Azure / GCP | Large model training | ~$3.00-5.00/hr |
| H100 (80 GB) | AWS / Azure / GCP | LLM training (latest) | ~$8.00-12.00/hr |
| TPU v5p | GCP only | Large-scale training (best) | ~$4.00-8.00/hr |
| Inferentia2 | AWS only | Cheap inference | ~$0.15/hr (Inf2 instance) |
| T4 (16 GB) | All 3 | Budget inference | ~$0.50-0.75/hr |
| Feature | SageMaker (AWS) | Azure ML | Vertex AI (GCP) |
|---|---|---|---|
| Notebooks | SageMaker Studio | Azure ML Studio | Vertex AI Workbench |
| AutoML | Yes (all types) | Yes (designer + SDK) | Yes (best-in-class) |
| Model Registry | Yes | Yes | Yes |
| Model Serving | Real-time + Batch | Real-time + Batch | Real-time + Batch + Predictions |
| Pre-trained Models | JumpStart, Bedrock | Azure AI Gallery | Model Garden (400+) |
| GenAI Models | Claude, Titan, Cohere | GPT-4o, DALL-E, Whisper | Gemini, Imagen, Chirp |
| MLOps | Pipelines, Experiments | Pipelines, MLflow | Pipelines, Feature Store |
| Special Hardware | Trainium, Inferentia | ND H100 v5 | TPU v5p (unique) |
| Category | AWS | Azure | GCP |
|---|---|---|---|
| Data Warehouse | Redshift | Synapse Analytics | BigQuery (best) |
| Pricing Model | $0.25/TB stored + $0.25/TB scanned | DWU units (reserved) | $5/TB queried (storage free) |
| Data Lake | S3 + Glue + Athena | ADLS Gen2 + Synapse | GCS + Dataflow |
| Streaming | Kinesis ($0.015/shard/hr) | Event Hubs ($0.01/event hub) | Pub/Sub ($0.40-40/M msgs) |
| Orchestration | Step Functions / MWAA | Data Factory / Synapse | Cloud Composer / Workflows |
| ETL/ELT | Glue ($0.44/DPU/hr) | ADF ($0.25/Orchestration) | Dataflow (auto-scaling) |
| Service | Cost/TB | Query Speed | Best For |
|---|---|---|---|
| BigQuery | $5/TB queried | Seconds (columnar + distributed) | Ad-hoc analytics, best value |
| Redshift | $0.25/TB stored + scan | Fast (provisioned clusters) | Large structured data, ETL |
| Synapse | DWU-based (variable) | Moderate | Azure-native orgs, Office data |
| Athena | $5/TB scanned (S3) | Moderate (serverless) | Occasional queries on S3 data |
| Snowflake (3rd) | $2-3/TB credited | Fast (auto-scaling) | Multi-cloud data warehouse |
| Reason | Description | Example |
|---|---|---|
| Vendor lock-in avoidance | Avoid dependency on single provider | Use K8s on multiple clouds |
| Best-of-breed services | Pick best service from each cloud | BigQuery (GCP) + Bedrock (AWS) |
| Compliance / Data residency | Different regions per regulation | EU data on EU servers |
| Negotiation leverage | Better pricing through competition | Negotiate with AWS using Azure quote |
| Acquisition / Merger | Combined orgs on different clouds | Migrate or integrate over time |
| Tool | Type | Description |
|---|---|---|
| Terraform | IaC | Multi-cloud infrastructure provisioning (most popular) |
| Pulumi | IaC | Infrastructure as code using Python/TypeScript |
| Crossplane | Control Plane | Kubernetes-native multi-cloud resource management |
| HashiCorp Vault | Secrets | Multi-cloud secrets management |
| Datadog | Monitoring | Unified observability across clouds |
| ArgoCD | GitOps | Multi-cluster Kubernetes deployment |
| Service Type | AWS | Azure | GCP |
|---|---|---|---|
| Serverless NoSQL | DynamoDB | CosmosDB (Serverless) | Firestore |
| Serverless SQL | Aurora Serverless v2 | SQL Database Serverless | Spanner (no serverless yet) |
| Message Queue | SQS (queue) / SNS (pub-sub) | Service Bus / Event Grid | Pub/Sub |
| Event Streaming | Kinesis Data Streams | Event Hubs | Pub/Sub (streaming) |
| In-Memory Cache | ElastiCache (Redis/Memcached) | Azure Cache for Redis | Memorystore (Redis/Memcached) |
| Full-Text Search | OpenSearch Service | Azure AI Search | Cloud Search (limited) |
| Service Type | AWS | Azure | GCP |
|---|---|---|---|
| Secrets Management | Secrets Manager + KMS | Key Vault | Secret Manager + Cloud KMS |
| Monitoring / Logging | CloudWatch | Azure Monitor + Log Analytics | Cloud Monitoring + Logging |
| Alerting | CloudWatch Alarms + SNS | Azure Alerts + Action Groups | Cloud Monitoring Alerting |
| Config / IaC | CloudFormation + CDK | ARM/Bicep + Terraform | Deployment Manager + Terraform |
| CI/CD Native | CodePipeline + CodeBuild | Azure Pipelines | Cloud Build |
| Container Registry | ECR | ACR | Artifact Registry |
| Phase | Activities | Tools |
|---|---|---|
| 1. Assess | TCO analysis, dependency mapping, readiness | AWS MAP, Azure Migrate, GCP Assessment |
| 2. Plan | Choose 6 Rs strategy, set timeline, budget | AWS Migration Hub, Azure Migrate |
| 3. Migrate | Data transfer, app migration, parallel run | AWS DMS/MGN, Azure Migrate, Google Transfer |
| 4. Optimize | Rightsizing, reserved instances, monitoring | Trusted Advisor, Azure Advisor, Recommender |
| 5. Operate | Ongoing management, cost control, updates | AWS SSM, Azure Arc, GCP Ops Agent |
| Task | AWS | Azure | GCP |
|---|---|---|---|
| Server migration | MGN (Application Migration) | Azure Migrate: Server Migration | Migrate to VMs (Velostrata) |
| Database migration | DMS (Database Migration) | Azure Database Migration | Database Migration Service |
| Data transfer (online) | DataSync | AzCopy / Storage Explorer | Storage Transfer Service |
| Data transfer (offline) | Snowball (50 TB) | Data Box (100 TB) | Transfer Appliance (480 TB) |
| VM discovery | Agentless Discovery | Azure Migrate appliance | Migration Center |
| Feature | AWS IAM | Azure RBAC + Entra ID | GCP IAM |
|---|---|---|---|
| Policies | JSON documents | Role definitions (JSON) | YAML/JSON bindings |
| Roles | Predefined + custom | Built-in + custom | Predefined + custom |
| MFA | Yes (virtual + hardware) | Yes (Entra ID) | Yes (Security Keys) |
| Federation | SAML 2.0, OIDC | SAML, OIDC, OAuth 2.0 | SAML, OIDC, Workforce/Workload |
| Service Accounts | IAM Roles for EC2/Lambda | Managed Identities | Service Accounts (K8s-style) |
| Conditional Access | IAM Conditions | Entra ID Conditional Access | IAM Conditions |
| Secrets | Secrets Manager + KMS | Key Vault | Secret Manager + Cloud KMS |
| Feature | AWS | Azure | GCP |
|---|---|---|---|
| Virtual Network | VPC (subnets, NACLs, SGs) | VNet (subnets, NSGs, ASGs) | VPC (subnets, firewall rules) |
| DDoS Protection | Shield (Standard free, Advanced $3K/mo) | DDoS Protection Standard | Cloud Armor (free tier) |
| WAF | WAF ($5/rule/mo + $0.60/M req) | WAF ($40/policy/mo) | Cloud Armor ($5/policy/mo) |
| Encryption at Rest | KMS (SSE-S3, SSE-KMS, SSE-C) | Azure Storage encryption | CMEK (customer-managed keys) |
| Encryption in Transit | TLS 1.2+ (enforced) | TLS 1.2+ (enforced) | TLS 1.2+ (enforced) |
| Threat Detection | GuardDuty ($0.35/GB scanned) | Defender for Cloud | Security Command Center |
| Metric | AWS | Azure | GCP |
|---|---|---|---|
| Total Regions | 34 regions | 60+ regions | 40 regions |
| US Regions | us-east-1, us-west-2, etc. | East US, West US, etc. | us-central1, us-east4, etc. |
| EU Regions | eu-west-1, eu-central-1, etc. | West Europe, North Europe | europe-west1, europe-west4 |
| Asia Regions | ap-southeast-1, ap-northeast-1 | East Asia, Southeast Asia | asia-east1, asia-southeast1 |
| CDN Edge Locations | 600+ | 200+ | 150+ |
| Sovereign Cloud | GovCloud, Secret Region | Azure Government, China 21Vianet | Air-gapped regions |
| Consideration | Guideline |
|---|---|
| Data residency laws | GDPR (EU), CCPA (CA), LGPD (Brazil), PIPA (Korea) — keep data local |
| Latency target | <100ms for interactive apps, <10ms for real-time/gaming |
| Same-region | Always co-locate compute, DB, and storage in the same region |
| Multi-region | Use for DR: async replication with <15 min RPO for most apps |
| Edge/CDN | Use CloudFront/Front Door/Cloud CDN for static assets — 10-50ms globally |
| Sovereign clouds | AWS GovCloud, Azure China (21Vianet), Google Air-gapped for gov |
| I Need To... | Recommended Cloud | Why |
|---|---|---|
| Run .NET apps with AD integration | Azure | Native .NET, Entra ID, Office 365 integration |
| Build an MVP cheaply | GCP | e2-micro free forever, Cloud Run scale-to-zero, BigQuery free tier |
| Train large ML models | GCP (TPUs) or AWS (P5) | TPUs for speed, AWS P5 (H100) for ecosystem |
| Use GPT-4 in production | Azure | Azure OpenAI is the only enterprise GPT-4 offering |
| Maximum service breadth | AWS | 200+ services, most mature, largest ecosystem |
| Run Kubernetes at scale | GCP (GKE) | Best-managed K8s — Autopilot, 99.95% SLA, free tier |
| Build a data warehouse | GCP (BigQuery) | Best pricing ($5/TB queried), separate storage/compute |
| Process streaming data | GCP (Pub/Sub) or AWS (Kinesis) | Pub/Sub for scale, Kinesis for AWS-native integration |
| Host static websites | Any (AWS S3, Azure Blob, GCP Cloud Storage) | All support static hosting + CDN |
| Enterprise compliance | Azure or AWS | Most certifications, FedRAMP High (AWS), broadest合规 |
| Factor | AWS | Azure | GCP |
|---|---|---|---|
| Service Breadth | ★★★★★ | ★★★★☆ | ★★★★☆ |
| Enterprise Readiness | ★★★★★ | ★★★★★ | ★★★★☆ |
| AI/ML Capabilities | ★★★★☆ | ★★★★☆ | ★★★★★ |
| Data & Analytics | ★★★★☆ | ★★★★☆ | ★★★★★ |
| Cost Effectiveness | ★★★☆☆ | ★★★☆☆ | ★★★★☆ |
| Kubernetes | ★★★★☆ | ★★★★☆ | ★★★★★ |
| Developer Experience | ★★★★☆ | ★★★★☆ | ★★★★★ |
| Startup Friendly | ★★★★☆ | ★★★★☆ | ★★★★★ |
| Cloud | Avoid When... | Consider Instead |
|---|---|---|
| AWS | You use Microsoft stack exclusively | Azure |
| AWS | You need cheapest always-free tier | GCP |
| Azure | You need best data analytics pricing | GCP BigQuery |
| Azure | You need TPU access for ML training | GCP |
| GCP | You need broadest service ecosystem | AWS |
| GCP | You need FedRAMP High certification | AWS or Azure |
| Certification | Level | Cost | Focus |
|---|---|---|---|
| Solutions Architect Associate | Associate | $300 | Designing distributed systems on AWS |
| Developer Associate | Associate | $300 | Developing and deploying AWS applications |
| SysOps Administrator | Associate | $300 | Deploying, managing, operating AWS workloads |
| DevOps Engineer Professional | Professional | $300 | CI/CD, automation, monitoring on AWS |
| Solutions Architect Professional | Professional | $300 | Complex solution design (multi-account) |
| Security Specialty | Specialty | $300 | AWS security architecture and operations |
| Data Analytics Specialty | Specialty | $300 | AWS data services (Athena, Glue, Redshift) |
| Certification | Provider | Cost | Focus |
|---|---|---|---|
| AZ-104: Administrator | Azure | $165 | Managing Azure identities, storage, compute |
| AZ-204: Developer | Azure | $165 | Developing Azure compute and storage solutions |
| AZ-305: Architect | Azure | $330 | Designing Azure infrastructure solutions |
| AI-102: AI Engineer | Azure | $165 | Building AI solutions with Azure AI services |
| DP-203: Data Engineer | Azure | $165 | Implementing data solutions on Azure |
| GCP Cloud Engineer | GCP | $200 | Deploying apps, monitoring, operations |
| GCP Cloud Architect | GCP | $200 | Designing GCP architecture (professional) |
| GCP Data Engineer | GCP | $200 | Designing data processing systems on GCP |
| Provider | Resource | URL / Access | Cost |
|---|---|---|---|
| AWS | Skill Builder | aws.amazon.com/training/ | Free (100+ courses) |
| AWS | Workshops | workshops.aws | Free hands-on labs |
| Azure | Microsoft Learn | learn.microsoft.com | Free (800+ modules) |
| Azure | Azure Sandbox | learn.microsoft.com/sandbox | Free limited-time access |
| GCP | Google Cloud Skills Boost | cloud.google.com/training | Free (select courses) |
| GCP | Qwiklabs | qwiklabs.com | Free credits + paid labs |
| Multi | A Cloud Guru / Pluralsight | acloudguru.com | Subscription ($39/mo) |
| Multi | KodeKloud | kodekloud.com | Free tier + subscription |