Managed Disks
Managed Disks are block-storage volumes that Azure manages as first-class resources and attaches to VMs. You pick a tier and a size; Azure handles the placement, replication, and availability of the underlying storage. Every VM boots from a managed OS disk and can attach managed data disks.
Disk tier is the most common quiet performance bug in Azure. A VM with plenty of vCPUs will still crawl if its disk caps IOPS at a few hundred. Block storage performance is provisioned, not magical, and the tier — not the VM size — sets the ceiling.
Disk Tiers
Tiers run from Standard HDD (cheap, low IOPS, for dev and archival) through Standard SSD (light production) to Premium SSD (production with consistent low latency). Premium SSD v2 and Ultra Disk decouple capacity from performance — you provision IOPS and throughput independently of size — for the most demanding databases.
| Tier | Latency | Use for |
|---|---|---|
| Standard HDD | High | Dev/test, backups |
| Standard SSD | Moderate | Light production, web servers |
| Premium SSD | Low, consistent | Production databases, latency-sensitive |
| Premium SSD v2 / Ultra | Lowest, tunable | High-IOPS databases, SAP HANA |
Performance
On the fixed tiers, IOPS and throughput scale with disk size — a larger Premium SSD is faster — and smaller disks can burst above their baseline for short spikes. On Premium SSD v2 and Ultra, you set IOPS and throughput directly, which is what lets a small-capacity volume serve a high-IOPS database without overpaying for unused gigabytes.
Snapshots and Images
A snapshot is a point-in-time copy of a disk, used for backups and to clone environments; an image captures a generalized VM for repeated deployment. Snapshots are incremental and cheap to keep, but a snapshot is not a backup strategy on its own — it lives in the same subscription and needs a retention and off-region plan, which is Azure Backup's job.
Encryption
Disks are encrypted at rest by default with platform-managed keys (server-side encryption). Customer-managed keys in Key Vault give you control of the key lifecycle for compliance. Encryption at host extends encryption to the temporary disk and the data in transit between the VM and storage, with no performance penalty.
Shared Disks
A shared disk can attach to multiple VMs at once, for clustered applications that coordinate access through their own cluster manager — Windows Server Failover Clustering, for example. It is a niche feature: without a cluster-aware filesystem on top, two VMs writing the same disk corrupt it.
Premium SSD — Performance scales with size; bursting for spikes. The production default for most latency-sensitive workloads.
Premium SSD v2 — IOPS and throughput provisioned independently of capacity. Cost-efficient for high-IOPS workloads that do not need much space.
Ultra Disk — The highest, fully tunable IOPS and throughput. For the most demanding databases such as SAP HANA — at the highest price.
- Running a production database on Standard HDD and blaming the VM size when the disk caps IOPS in the hundreds.
- Provisioning a huge Premium SSD purely to get IOPS, when Premium SSD v2 would deliver the IOPS at a fraction of the capacity cost.
- Treating snapshots as a backup strategy — they live in the same subscription with no retention policy or off-region copy.
- Attaching a shared disk to multiple VMs without a cluster-aware filesystem, corrupting it with concurrent writes.
- Forgetting that detached disks still bill — orphaned data disks from deleted VMs quietly accumulate cost.
- Skipping customer-managed keys where compliance requires control of the key lifecycle, then discovering it during an audit.
- Use Premium SSD for any production workload with a latency requirement; reserve Standard tiers for dev and archival.
- Use Premium SSD v2 or Ultra when you need high IOPS without the capacity a fixed-tier disk would force you to buy.
- Pair snapshots with Azure Backup for real retention and off-region protection.
- Enable encryption at host for full-path encryption with no performance cost; use customer-managed keys where compliance demands it.
- Clean up orphaned disks from deleted VMs — detached disks keep billing.
- Size fixed-tier disks for the IOPS you need, remembering that performance scales with capacity.
Knowledge Check
A production database VM is slow despite ample vCPUs and memory. The disk is Standard HDD. What is the cause?
- The disk tier caps IOPS and latency regardless of the VM size
- The VM needs more vCPUs allocated to it to drive the slow disk
- Standard HDD encrypts every read on the fly, which slows the database down
- The disk must be configured as a shared disk to perform well
When does Premium SSD v2 beat a fixed-tier Premium SSD?
- When you need high IOPS but little capacity — v2 provisions IOPS independently of size
- When you need the absolute lowest per-GB storage cost for rarely accessed cold archival data
- When attaching one disk to many VMs at once for clustering
- When the workload is dev or test with no real performance need
Why is a disk snapshot not a complete backup strategy?
- It lives in the same subscription with no retention policy or off-region copy
- A snapshot can never be restored or used to create a brand-new managed disk later
- Snapshots are stored unencrypted and readable as plain text
- Taking a snapshot deletes the source disk in the process
You got correct