Put time in your plan.
Measure the pace first.
Model an operation or sequential restore chain using observed throughput—not a disk’s advertised speed.
Operation profile
01 / INPUTSDuration plan
02 / RESULTSBaseline duration · data processing + overhead
Planning summary
Read-only progress query
SELECT session_id, command, percent_complete,
total_elapsed_time / 60000.0 AS elapsed_minutes,
estimated_completion_time / 60000.0 AS estimated_remaining_minutes
FROM sys.dm_exec_requests
WHERE command LIKE 'BACKUP%'
OR command LIKE 'RESTORE%';Live estimates can change and may be unavailable. Requires VIEW SERVER STATE, or VIEW SERVER PERFORMANCE STATE on SQL Server 2022. This page does not connect to SQL Server.
A rate is useful
only with matching bytes.
Backup and restore run through different bottlenecks. Measure them separately on representative hardware and workloads.
Formula and units
Data-processing seconds = total GiB × 1,024 ÷ measured MiB/second. Baseline duration adds your fixed overhead. The faster scenario uses rate × (1 + variation); the slower scenario uses rate × (1 − variation). Fixed overhead remains unchanged.
GiB and MiB use binary units. If your measurement is decimal MB/s, multiply it by 1,000,000 ÷ 1,048,576 before entering it. Defaults are examples, not a measured benchmark.
Choose one measurement basis
Logical backup data pairs with a logical processing rate. Stored bytes pair with measured backup-device read/write throughput. Compression affects the relation between these quantities; do not divide again by a compression ratio when using an observed effective rate.
Restore-chain planning
For a chain, enter the sum of the full, selected differential and required log sizes using the same basis. The tool assumes sequential processing at one effective rate. Different stages can have very different rates; estimate each separately when needed.
Restore can also require file initialization, redo and undo. Add those measured costs as overhead only if they are not already included in your rate. Retrieval, decryption, validation, retries, application cutover and database checks may take additional time. A modeled window comparison is not an RTO guarantee.
Guidance checked October 5, 2026. Test representative restores to establish actual recovery duration.