What is a backup window?

A backup window is the scheduled period when backup jobs run, chosen to limit the load on production systems, the network and storage. It is usually overnight or at weekends. The amount of data that fits depends on the slowest point between source and target, and the window stays fixed while data grows.

What limits how much fits in the window?

A backup is a pipeline: the source systems, the backup server or proxy that reads and processes the data, the network, and the backup target. The job runs at the speed of the slowest of these. Duration is roughly data volume divided by that sustained rate: 10 TB at 500 MB per second takes about five and a half hours, and the same data at half the rate takes about eleven. Upgrading the network changes nothing if the limit is the proxy's processing, the source's read rate or the target's ingest rate. The bottleneck is found by measuring each segment while a job runs.

What load does a backup put on the source?

Reading data for a backup consumes disk reads, CPU and memory on the source systems, and snapshots held open during the job add overhead to production storage. Databases and busy virtual machines slow down noticeably while they are read in full. That source load is the main reason the window sits outside business hours, and it is why a job that overruns into the morning affects users, not only the backup team.

Why do windows overrun over time?

The window does not grow, and data does. A job that once finished in the early hours reaches the start of the working day without any setting having changed. Overruns lead to skipped or partial jobs, and each skipped night widens the gap between restore points, which increases the data that could be lost in a recovery and puts the recovery point objective at risk.

What shares the window, and what reduces the load?

Full backups, incrementals, off-site copy jobs, verification and storage maintenance all draw on the same hours and often the same components. Several methods reduce the load. Incremental backups and change tracking limit daily jobs to changed data. A synthetic full backup builds a new full on the backup storage from existing copies, so production is not read again. Spreading jobs across the window and across proxies lowers the peak demand on any one component.

Frequently asked questions

Do continuous backup methods remove the window?

They spread the load across the day by copying changes as they occur. A window still exists for tasks such as full copies and verification that cannot run continuously.

Does a longer window fix overruns?

Only for a time. As data grows the longer window fills too, and a window that extends into business hours moves the source load onto users.

Is the backup window the same as the recovery time?

No. The window is the time allowed to take backups. Recovery time is how long a restore takes, which often runs at a different and lower rate.