Optimize Azure managed disk costs

You pay for what is provisioned

Managed disks are generally billed on provisioned size and performance characteristics rather than on the data actually written. A 1 TB disk holding 80 GB usually costs the same as a full one. That makes provisioning decisions, not cleanup inside the guest, the main cost lever.

Match the tier to measured demand

Disk families differ in latency, throughput and IOPS behaviour. Some workloads need consistent low latency; others are largely idle. Review disk metrics over a representative period, including peaks and scheduled jobs, before changing tier. Changing a disk type can require the VM to be stopped, and some conversions are restricted.

1. Collect disk IOPS, throughput and latency for a representative window.
2. Compare observed demand with the limits of the candidate tier.
3. Check VM-level limits, which can cap disk performance regardless of tier.
4. Plan a maintenance window and a rollback option before converting.

Snapshots and images accumulate quietly

Snapshots are separate billable resources. Automated processes that snapshot before every change can build a long tail of forgotten copies. Review snapshot inventory and age, confirm retention requirements, and remove only what an owner agrees is no longer needed.

Unattached disks are the obvious win — with a caveat

Disks left after a VM is deleted keep billing. Many are safe to remove, but some are retained deliberately for recovery or forensics. Confirm ownership and check for resource locks, then consider a snapshot-and-delete approach where policy allows — remembering the snapshot itself has a cost.

Measure the result properly

Illustrative example: removing four unattached disks that cost ₹500 each per month suggests around ₹2,000 per month less spend. Verify against the next complete billing period, since mid-month deletion leaves partial charges. Cloudledger’s Orphaned Resources view helps build the candidate list with billing evidence.