Azure Compute Architecture Hub 🖥️⚡
Microsoft Azure provides a comprehensive compute continuum spanning raw infrastructure-as-a-service (IaaS) virtual machines, managed serverless application runtimes, elastic container platforms, and hyperscale GPU clusters for artificial intelligence.
Architectural Decision Framework
Selecting the optimal Azure compute platform requires balancing infrastructure control, operational overhead, cold-start latency, and scaling dynamics:
Compute Requirement
│
┌──────────────────────────────┴──────────────────────────────┐
▼ ▼
Containers / Cloud-Native Legacy / Full OS / Monolith
│ │
┌───────┴───────┐ ┌───────┴───────┐
▼ ▼ ▼ ▼
Event-Driven Microservices / Standard App Scale & Elastic
Small Units Full Kubernetes Single VM Fault Domains
│ │ │ │
▼ ▼ ▼ ▼
[[Azure/compute/functions|Azure Functions]] ┌──────┴──────┐ [[Azure/compute/vm|Azure VM]] [[Azure/compute/vm|VMSS]]
│ │
Low Complexity Full Control
│ │
▼ ▼
[[Azure/compute/container-apps/README|Azure Container Apps]] [[Azure/compute/aks|Azure Kubernetes Service (AKS)]]
Compute Platform Comparison Matrix
| Service | Abstraction Level | Scaling Mechanism | Networking & Isolation | Primary Workload Fit |
|---|---|---|---|---|
| Virtual Machines (VM) & VMSS | Full IaaS (Linux/Windows OS) | Horizontal Autoscale (VMSS metrics/schedule) | Direct VNet NIC, Availability Zones, Proximity Placement | Legacy monolithic apps, stateful clustering, custom kernel drivers, HPC. |
| Azure App Service | PaaS (Web & API Hosting) | Scale across App Service Plan instances | Regional VNet integration, Private Endpoints, Deployment Slots | Web apps, REST APIs, enterprise web tiers needing zero-downtime blue/green slots. |
| Azure Container Apps (ACA) | Serverless Containers (Managed Envoy/KEDA) | Event-driven KEDA scalers (Scale to Zero) | Managed VNet environment, internal service discovery, Dapr | Microservices, background queue workers, event-driven containerized APIs. |
| Azure Kubernetes Service (AKS) | Managed Kubernetes Engine | Cluster Autoscaler, Karpenter (NAP), HPA, KEDA | Azure CNI, CNI Overlay, Cilium eBPF, Pod Subnets | Complex multi-tier microservices, distributed stateful systems, high-density fleets. |
| Azure Functions | Serverless FaaS (Event-driven) | Scale Controller, Flex Consumption (up to 1,000) | Regional VNet integration, Private Endpoints | Webhooks, stream consumers, asynchronous ETL, event-driven reactive micro-tasks. |
Service Catalog
1. IaaS & Virtual Machines
- Virtual Machines & Scale Sets (VMSS) — VM series taxonomy (B, D, E, F, M, N), Flexible VMSS orchestration, Availability Zones vs Fault Domains, and Proximity Placement Groups.
- Azure Spot VMs & Scheduled Events — Surplus capacity at up to 90% discount, 30-second eviction handling via Scheduled Events API, and fault-tolerant batch architectures.
2. PaaS & Serverless Application Platforms
- Azure App Service & Deployment Slots — Dedicated App Service Plans, zero-downtime deployment slots with traffic routing, custom domain TLS, and regional VNet integration.
- Azure Container Apps (ACA), KEDA & Dapr — Serverless containerized microservices built on AKS/Envoy, event-driven KEDA autoscaling, Dapr microservice sidecars, and scale-to-zero semantics.
- Azure Functions & Serverless Computing — Event-driven FaaS, Flex Consumption, Isolated Worker Model, Durable Functions stateful orchestrations, and cold-start optimization.
3. Managed Kubernetes (AKS Ecosystem)
- AKS Master Architecture Hub — Complete engineering guide indexing 22 specialized AKS deep dives:
- AKS Automatic — Fully managed Kubernetes, Karpenter-powered Node Auto-Provisioning (NAP), automated weekly maintenance.
- Cluster Tiers & SLA — Free vs Standard (99.95% SLA) vs Premium (LTS), Private Link clusters, API Server VNet Integration.
- Networking & CNI Overlay — Kubenet vs Azure CNI vs Azure CNI Overlay vs Dynamic Pod IP Allocation.
- Azure CNI Powered by Cilium — eBPF kernel datapath, kube-proxy replacement, WireGuard encryption, Hubble observability.
- Ingress & Gateway API — Application Gateway for Containers (AGfC), Gateway API (
HTTPRoute), and WAF v2. - Heterogeneous Node Pools — System vs User pools, Ephemeral OS disks on NVMe/Cache, Azure Linux 3, and Spot pools.
- Autoscaling Architecture — Cluster Autoscaler (CA), Node Auto-Provisioning (Karpenter), HPA v2, KEDA, and Virtual Nodes.
- ML Infrastructure — NVIDIA H100/H200 (NDv5), Quantum-2 3.2 Tbps InfiniBand RDMA, KubeRay, and vLLM serving.
- Storage CSI (Disks) — Azure Managed Disks, Premium SSD v2, Ultra Disk, and Azure Elastic SAN.
- Shared Storage CSI (Files & Blob) — Multi-pod RWX, Azure Files NFS v4.1, and Azure Blob Storage BlobFuse2.
- Entra Workload Identity — Keyless authentication, OIDC federated credentials, and Azure RBAC for Kubernetes.
- Key Vault Secrets Store CSI — In-memory tmpfs mounts, secret auto-rotation, and FIPS 140-2 Level 3 HSM.
- Azure Policy & Gatekeeper — OPA Gatekeeper admission guardrails, CIS benchmarks, and custom Rego constraints.
- Multi-Tenancy & Confidential Containers — PSS, Azure Linux Kata Containers (Hyper-V micro-VMs), and AMD SEV-SNP.
- Observability & Monitoring — Azure Monitor Container Insights, Managed Prometheus, Grafana, and ContainerLogV2.
- Backup & Disaster Recovery — Azure Backup for AKS (managed Velero), BackupVault, and multi-region Active-Active Front Door.
- SRE Troubleshooting Runbook — Exit code taxonomy (137 OOMKilled), Node NotReady triage, and CNI IP exhaustion.
- FinOps & Cost Allocation — Microsoft Cost Management AKS Cost Allocation, Savings Plans, and pause pod overprovisioning.
- Azure Kubernetes Fleet Manager — Staged rolling update runs across clusters and Multi-Cluster Services (MCS).
Multi-Cloud Equivalents
| Capability | Microsoft Azure | Amazon Web Services (AWS) | Google Cloud Platform (GCP) |
|---|---|---|---|
| Virtual Machines | Azure Virtual Machines | Amazon EC2 | Compute Engine |
| Scale Sets | Virtual Machine Scale Sets (VMSS) | EC2 Auto Scaling Groups (ASG) | Managed Instance Groups (MIG) |
| Web PaaS | Azure App Service | AWS Elastic Beanstalk / App Runner | App Engine |
| Serverless Containers | Azure Container Apps (ACA) | AWS App Runner / ECS Fargate | Cloud Run |
| Managed Kubernetes | Azure Kubernetes Service (AKS) | Amazon EKS | Google Kubernetes Engine (GKE) |
| Serverless Functions | Azure Functions | AWS Lambda | Cloud Functions / Cloud Run functions |