Interpret S2D drive history at the device measurement boundary

Source facts

Microsoft provides drive history for devices in the cluster storage subsystem, excluding operating-system boot drives. The IOPS, throughput, and latency series come from the connected server’s Physical Disk counters, measured by partmgr.sys. They exclude much of the Windows software stack and network traversal. The values cover the whole interval: Microsoft’s example averages thirty operations in a ten-second interval as three operations per second. Microsoft documentation.

Applicability

Identify the physical drive, connected server, metric, time interval, and application event. Distinguish the device-level observation from an end-to-end application response measurement before attributing a delay.

DSE recommendation

Record the drive identity and measurement boundary alongside each performance finding. Correlate device observations with workload and network evidence from the same period. Ask the investigator to state whether a value is an interval average or another aggregation before comparing it with a threshold.

Verification

Check the requested drive’s series and timestamps and compare them with the planned measurement interval. Investigate a high device latency alongside the remaining I/O path rather than assigning the whole application delay to that drive. Preserve missing history or an excluded boot device as a scope limitation.

Official references

Microsoft Learn: Performance history for drives. Source reviewed September 8, 2026.