For businesses, having a data backup strategy is crucial, but it is often overlooked. Including it in post-provisioning tasks can save you time and effort when rebuilding a server. When handing over a server to another team, an installation can go wrong, and having a backup or snapshot can make it easy to restore the server to the last good state, which can be a lifesaver.

To determine backup needs, it is important to firstly understand the business requirements that drive them. Are there long data retention needs, or is quick recovery more important? Are there any regulatory compliance requirements or client service level agreements that dictate restore times or retention periods? For businesses operating in Vietnam, you can refer to Decree 53 on Cybersecurity Law to better understand these requirements.

The choice of backup methods depends entirely on the business needs. Therefore, businesses can start selecting suitable methods by understanding the business requirements and then identifying which option meets those needs.

1. Snapshot/Redirect-on-write

Using redirect-on-write snapshots is a clever technology that can reduce the performance impact of snapshots compared to copy-on-write and other methods. Essentially, the storage provider maintains a table that lists the location of various blocks, and each location is referenced by a pointer in the table, which points to the location without actually containing the data. When a write operation is performed and the block needs to be modified, the pointer is simply updated to the new location where the data is modified, rather than copying the original block, modifying the copy, and then overwriting it.

One excellent application of this technology is to create a snapshot prior to an upgrade, whether it's an operating system upgrade, a software update, or an application package update. This enables a quick rollback in the event of an issue during or shortly after the upgrade. Snapshots typically revert quickly and take up minimal space, but the longer they exist, the more data they will consume as more changes occur. Some environments are more suitable for long-standing snapshots than others. For instance, some hypervisors advise against keeping a hypervisor snapshot for more than a few days, while cloud providers and many SAN (Storage Area Network) vendors don't have any such performance issues.

2. Clone

Cloning is a backup method for several use cases, primarily centered around duplicating systems for quick recovery or spinning up new instances. However, this method typically consumes a lot of storage since it creates an identical copy, including metadata, for faster restoration. Essentially, cloning is both a backup and restore operation in one.

Commonly, cloning is used to duplicate virtual machines for creating multiple instances. In some cases, it serves as a temporary backup for quick restoration in case of any issues with the source. One practical application of cloning is to replicate a production server in a sandbox environment to test upgrades. Lower-level DEV, QA, and UAT systems can successfully complete a specific upgrade. However, peculiarities may only exist in production, and this use case can aid in testing scenarios that are unique in production.

3. Full Backup

Full backup is the traditional and widely-used full backup that involves creating an identical copy of the data, similar to a clone. However, the storage and location of the backup can vary. For example, a clone may be stored alongside the original in the same location and format, whereas a backup is typically stored in the backup software's native format to enable compression and deduplication. Additionally, it's usually saved on a different medium, such as local servers, hard drives, flash drives, and CDs.

System administrators often feel relieved after performing a full backup because it provides a last known good restore point in case of any issues.

4. Differential Backup

A differential backup is a type of backup that captures and stores changes made since the last full backup. It is faster than a full backup since it only saves the changes, but during a restore, the full backup must be restored first, followed by the differential backup. The main difference between a differential backup and an incremental backup is that the former only checks for changes made since the last full backup, and not since any other backup type.

Differential backups are often used throughout the week to reduce the number of restores required to reach the latest data. More information on this will be discussed in the incremental backup section.

5. Incremental Backup

Incremental backups differ from differential backups in that they track changes since the last backup, whether it is a full, differential, or incremental backup. For instance, if a full backup is done on Sunday, and incremental backups are done until the next Sunday, a restore would require the full backup and every incremental backup to the point of the restore.

This method provides a faster backup, but restoring can take longer. The choice between Incremental and Differential backups depends on the priority: decreased server load during backups and shorter backup windows or quicker restores. When considering databases, a backup schedule with more differentials may be useful for achieving restore time service level agreements (SLAs), while incrementals may be more helpful for reducing backup loads if the SLA for restoring data is relaxed.

6. Change Block/Delta Tracking

Change block or delta tracking is a widely used technique that facilitates incremental or differential backups by supporting to identify what has changed. Instead, the backup software can simply access the list of modified items and start backing them up. Without this feature, the software would have to scan the data to determine whether it has already been backed up.

This is not a distinct type of backup, but rather a backup feature that can enhance the performance of incremental or differential backups. Many backup types can benefit from this feature, as it can significantly reduce backup times. However, the underlying storage provider must support change block or delta tracking in order for this feature to be available.