Picking the right Azure managed disk tier means matching the disk's performance class to the IOPS and throughput a workload genuinely uses, not to the size of the VM it is attached to. Azure offers five managed disk types that trade price against performance: Ultra Disk, Premium SSD v2, Premium SSD, Standard SSD, and Standard HDD. Most over-spending comes from leaving every disk on Premium SSD by default when a large share of them never see the load to justify it.
This article is part of our Azure cluster. Start with the complete guide to Azure cost optimization, the pillar this piece links up to. Disk tiering is a Cut step in our See, Cut, Lock, Run method, applied to the block storage layer that sits underneath every VM.
When a VM is created, the OS disk and many data disks land on Premium SSD because that is the comfortable default. Premium is the right call for production databases and latency-sensitive workloads. It is the wrong call for dev boxes, batch nodes, log volumes, and archival data that are read occasionally. The fix is not heroic: it is reading the actual disk metrics and stepping each disk down to the cheapest tier that still meets its real demand.
The five disk tiers and what each is for
Azure managed disks span a clear performance ladder. The colder you go, the cheaper the per-GB storage rate, and the lower the guaranteed IOPS and throughput. The skill is reading where each workload actually sits on that ladder rather than assuming it needs the top of it.
| Disk type | Best for | Watch out for |
|---|---|---|
| Ultra Disk | Top-tier databases, SAP HANA, workloads needing very high IOPS and sub-millisecond latency with tunable performance | Most expensive; performance you configure and pay for even if unused |
| Premium SSD v2 | Production workloads wanting high performance with IOPS and throughput provisioned independently of capacity | Newer model; confirm regional availability and feature support |
| Premium SSD | Production VMs, transactional databases, latency-sensitive apps | Default that gets over-applied to non-production disks |
| Standard SSD | Web servers, lightly used enterprise apps, dev and test | Lower, less consistent performance than Premium |
| Standard HDD | Backup targets, infrequently accessed data, non-critical batch | Lowest and most variable performance; not for latency-sensitive work |
Step 1: Read the disk metrics before you touch anything
The mistake is changing tiers from a hunch. Pull the per-disk IOPS and throughput from Azure Monitor over a representative window, ideally including a busy period like month-end or a peak traffic day. A disk that peaks at a few hundred IOPS on Premium SSD is paying for performance it never consumes and is a clean candidate for Standard SSD. A disk that sustains high IOPS and is latency-sensitive belongs where it is. Decisions made on real metrics hold up; decisions made on instance size do not.
Step 2: Separate OS disks from data disks
Treat them differently. OS disks rarely need Premium performance on non-production machines, so a fleet of dev and test VMs can often run Standard SSD OS disks with no noticeable effect. Data disks are where the real performance need lives or does not, so size each one to its workload rather than inheriting the VM's default. On large fleets, the OS-disk downgrade alone returns meaningful savings because it repeats across every machine.
A fleet of Premium disks nobody profiled?
Our Azure cost audit reads the IOPS and throughput on every managed disk, flags the ones over-provisioned for their real load, and quantifies the saving from right-tiering before anything changes. On the performance model, you pay only from realized savings. No savings, no fee.
Book an Azure cost audit →Step 3: Understand how each tier bills
The tiers do not all charge the same way, and that shapes the decision. Premium SSD and Standard SSD bill largely on the provisioned disk size, so capacity drives the cost. Premium SSD v2 and Ultra Disk decouple capacity from performance: you provision storage, IOPS, and throughput separately and pay for each. That decoupling is powerful when used deliberately, because you can buy exactly the performance a workload needs instead of buying it implicitly through a larger disk. It is also a trap if you over-provision the IOPS and throughput dials and forget them, because you keep paying for capability you never use.
Step 4: Right-tier, then re-check
Changing a disk tier is a low-drama operation for most disk types, but it is not free of consequence: confirm the workload tolerates the new performance profile, and watch latency and queue depth for a few days after the change. Right-tiering is not a one-time sweep either. New VMs keep landing on Premium by default, so the profiling needs to repeat, which is exactly why we fold it into governance rather than treating it as a project. Set the expectation that disks get profiled on a schedule and the saving stays in place.
Don't forget the orphans
Right-tiering only addresses disks that are attached to something. The other half of disk waste is disks attached to nothing: volumes left behind when a VM is deleted that keep billing every month. Sweep those as part of the wider hunt in how to find idle and orphaned Azure resources, and clean the snapshot sprawl that accumulates alongside them in how to clean up Azure snapshots and backups. An orphaned Premium disk is the most expensive kind of nothing.
Disk type names and the independent-provisioning behavior described above reflect Azure managed disks as of May 2026. Verify current disk families, performance limits, and pricing in Azure documentation before re-tiering production storage, as these change.
The Azure Cost Optimization Field Guide includes the disk-tier decision matrix and the Azure Monitor queries we use to profile IOPS and throughput on engagements. It is the downloadable companion to this article.
The short version
Match each disk to the cheapest tier that meets its real IOPS and throughput, profile from metrics rather than VM size, downgrade non-production OS disks across the fleet, and watch the performance dials on Ultra and Premium SSD v2 so you do not over-provision them. Then sweep the orphaned disks that bill for nothing. To place disk tiering inside a full pass, follow how to run an Azure cost optimization assessment. When you want every disk profiled and right-tiered across the estate, that is exactly what our Azure cost optimization service delivers.