Updated 26 October 2026 · Published by Zoorik
Disk latency is how long one read or write takes to finish. The clock runs from sending the request until the disk confirms it’s done. You’ll usually see it in milliseconds (ms).
It’s also called storage latency or input/output (I/O) latency. I/O operations per second (IOPS) counts the operations, and latency is the time each one takes.
Azure disk latency and EBS latency
Each Azure disk type has a design target:
- Standard HDD: writes under 10 ms and reads under 20 ms, for most operations.
- Standard SSD and Premium SSD: single-digit milliseconds.
- Premium SSD v2 and Ultra Disk: under 1 ms.
On AWS, CloudWatch reports average EBS latency per minute for volumes on Nitro-based instances, as VolumeAvgReadLatency and VolumeAvgWriteLatency.
There’s no single acceptable disk latency. Transaction-heavy apps, such as databases, are sensitive to it. Apps that stream large files care less.
Why it matters for your cloud cost
Disk types differ in latency and price. But high latency can also mean a disk or VM is at its limit. Check both limits before you change type:
- Ask a disk for more IOPS or throughput than it allows, and latency can rise.
- The VM has its own limits too.
Amoeba moves a VM’s data disks onto one pool sized to your data, then rightsizes it autonomously. A 2-hour soak follows the switch to the pool.
The p99 marks the time 99 in 100 operations finish within. During the soak, any disk error, or p99 disk latency above 1.2 times its level before the switch, pauses the move.
The free assessment works out your own figure.
Read next: Azure disk IOPS vs throughput: what limits your disks
Sources: AWS, “Amazon EBS I/O characteristics and monitoring” (opens in a new tab); AWS, “Amazon CloudWatch metrics for Amazon EBS” (opens in a new tab); Microsoft Learn, “Azure premium storage: Design for high performance” (opens in a new tab); Microsoft Learn, “Select a disk type for Azure IaaS VMs” (opens in a new tab).