Azure IaaS (VM) Best Practices: Are Your VMs Sized by Reality or Tradition?
Why Microsoft Azure Best Practices Matter
- Reliability is trust. Quiet VMs keep SLAs and calendars intact.
- Spend follows design. Rightsizing and scheduling beat "throw more CPU at it."
- Security lives at the edge. Every VM is a border; close the obvious gates, and drift slows down.
- Ops morale. Automanage + Update Manager reduces toil so your team ships improvements, not excuses.
Sizing by Telemetry
Are those cores working or just warming the bench?
Use P95 CPU/memory from VM Insights over 30+ days. Target approximately 30–60% CPU and 50–75% memory for steady workloads; scale out for spikes.
Plain talk: Assuming larger capacity is safer represents a comfortable but costly cloud mentality.
Gen2 Images & Image Discipline
Why bother with pipelines for VMs?
Gen2 unlocks Secure Boot/vTPM and compatibility. Centralize builds via Azure Image Builder and distribute with Shared Image Gallery so every VM isn't a special snowflake. Version images and roll back when needed.
Career tip: Quick rollback capabilities can resolve critical Friday incidents.
Availability & DR
What breaks first: zones or backups?
Assume failure, prove recovery. Spread across Availability Zones or VM Scale Sets (flex); test ASR restores quarterly with real RPO/RTO.
Reality check: An untested backup lacks practical validation.
Access & Secrets
Why is "no public RDP/SSH" still a debate?
It isn't. Use Bastion or Defender JIT. Swap hard-coded secrets for Managed Identities; store the unavoidable ones in Key Vault with purge protection.
Truth bomb: Shareable credentials represent a security vulnerability.
Operations That Stay Out of the Headlines
Can patching be painless?
Yes. Automanage and Update Manager standardize baselines, patch windows, and monitoring. Pair with start/stop schedules for non-prod, Reservations/Savings Plans for steady loads, and Spot for batch.
Financial sanity: Scheduling enables responsible resource management when operations conclude.