Microsoft Azure · Infrastruttura · YoctoIT tech page

AKS

Azure's managed Kubernetes: a worry-free control plane and native integration with corporate identity, monitoring and policies.

FOCUS · K8S IN THE MICROSOFT HOUSEEphemeral clusters, Entra identity and the right node pools: AKS done well
YoctoIT material for clients and partners · Microsoft, Azure, Entra, Defender and the other products mentioned are trademarks of Microsoft Corporation.
01 · What it is

Azure Kubernetes Service, made clear.

AKS removes the control-plane toil and integrates Kubernetes into the Microsoft world: access via Entra ID, telemetry in Azure Monitor, policies via Azure Policy, the ACR registry. What remains are the choices that make the difference: node pools, networking, autoscaling and supply-chain security.

Free
the managed control plane isn't billed: you pay only for the nodes
Entra ID
Kubernetes RBAC tied to the corporate identity: whoever gets in is who you say
KEDA
event-driven autoscaling: consumers scale on the queues, not on CPU
Azure Kubernetes Service
OFFICIAL MICROSOFT BRANDING · AKS
PORTALE AZURE REALE · CREAZIONE CLUSTER AKS · FONTE: MICROSOFT LEARN
REAL AZURE PORTAL · AKS CLUSTER CREATION · SOURCE: MICROSOFT LEARN
02 · How to use it well

The things that make the difference.

The typical architecture

Microservices & jobsthe containerized workloads
System node pool
User node pool
Spot pool
separated by role and cost
AKS · managed control planeMicrosoft's upgrades and SLA
VNet, ACR, Key Vault, Monitorthe ecosystem around it
Kubernetes, with the Azure ecosystem around it

Node pools with judgment

System separated from workloads, spot for batch, GPUs where needed: the bill and stability say thanks.

Networking chosen early

Azure CNI, private clusters, ingress: network choices are made at the start or paid for later.

Upgrades without anxiety

Release channels and maintenance windows: the updated cluster stays a boring event.

GitOps and policies

Integrated Flux and Azure Policy on the clusters: configuration governed as code.

03 · In depth

Node pools, network and governed upgrades

AKS separates system and user node pools (spot included), with per-pool autoscalers and maintenance windows for upgrades; releases follow channels (rapid/regular/stable) and LTS gives 2 years on the same minor. The network is a choice: Azure CNI overlay to save IPs or pod subnet for full integration; Workload Identity federates pods on Entra; deployment safeguards block out-of-standard manifests.

  • Node pool — separate system, user for workloads, spot for batch: clear lanes
  • LTS — 2 years on the same minor for those who can't chase K8s
  • Azure CNI overlay — virtual pod IPs: the corporate CIDR breathes
  • Workload Identity — pods authenticated on Entra: zero secrets in the cluster
  • Maintenance window — upgrades when you say so: no nightly surprises
  • Karpenter (NAP) — node auto-provisioning: the right size decided by the pods
04 · Numbers and lifecycle

The numbers that matter.

99,95%
control plane SLA with uptime SLA enabled
2
years of LTS for selected minors
3
the release channels: speed is a choice
5k
nodes per cluster: the scale is there
AKS must be governed as a fleet: pools, identity and channels set up by us — and upgrades become routine maintenance.
05 · Use cases

Where it really pays off.

Microservices platform

The standard runtime for corporate cloud-native apps.

Burst and batch

Jobs and peaks on spot nodes: power at clearance prices.

Hybrid with Arc

The same patterns on on-prem clusters too, governed from Azure.

Managed Kubernetes doesn't mean abandoned Kubernetes: design, security and 24/7 provided by us.