Data backup methods: full, incremental and differential backups

Considering the fact that antivirus software cannot catch up to the development of malware, the most rational choice is to build IT security around reserve copying systems. Instead of focusing solely on prevention, it is much easier and cheaper to recover your system from a backup. Moreover, reserve copying can mitigate the results of a force-majeure, human error or hardware failures. In this article, we will look into the advantages and disadvantages of the most common types of reserve copying.

What is a full backup?

A full backup creates a copy of the entire system every time, or copies of the files that you have chosen for reserve copying. The data is then archived to save space. This way, new archives appear in your storage periodically; most copies end up being redundant because the data remains unchanged since the last backup. This is a significant disadvantage because of ineffective resource usage: storage space, CPU time, the time it takes to create the copies, computing power and traffic resources used to transfer the archives. Full backups used to be common due to high reliability; nowadays, it is considered ineffective when used on its own. For example, full backups waste a lot of resources when the backup depth is low (less than two weeks) or if the frequency is high (once every 24 hours or once every couple of hours).

The dedupliction mechanism is often used to detect and delete copies of data in full backups. It is used both on the levels of the data storage or the server and the client itself. Deduplication can be quite effective: it can remove from 90% to 98% redundant copies. One of the few advantages of full backups is the recovery speed: when the data is recovered from a single archive, the process is faster than with incremental or differential backups. Nowadays, full reserve copying is most often used as the base, combined with other methods that use fewer resources. This data backup and recovery strategy is sometimes called a mixed or a synthetic backup.

What is an incremental backup?

This type of reserve copying is much faster and more resource-effective that a full backup, because it only copies the files that have changed since the last backup’s creation. The data that was uploaded initially is not overwritten. The mechanism of incremental copying is simple: point X0 is chosen and set up for a certain time (for example, midnight between Sunday and Monday) and a full backup is created; files that have been changed or created since point X0 are uploaded at point X1 (midnight between Monday and Tuesday); files that have been changed or created since point X1 are uploaded at point X2 (midnight between Tuesday and Wednesday)… the cycle ends at point Xn and a new full backup is created. This method uses fewer resources and less storage space compared to other backup types. However, data is recovered from points Хn-1…Х2, Х1, Х0 up until the last full backup, which can take plenty of time depending on the volume of data.

We have chosen this backup type for our own cloud reserve copying service – BaaS (Backup-as-a-Service). To solve the issue of recovery speed, we offer two options for clients – BaaS Local and BaaS Remote. Data is either saved in the same data center as the client’s infrastructure or on a remote platform. BaaS Local improves recovery time while BaaS remote increases data security.

What is a differential backup?

Incremental and differential backups differ in terms of recovery speed; full X0 and Xn copies are compared and there is no need for recovering data stage-by-stage. However, differential backups take up nearly as much space as full backups, which doesn’t solve the task of saving storage space.

Differential reserve copying is done on an accrual basis: each changed file in the next backup point is copied again. This can be represented as: Х0, Х1, Х1+Х2, Х1+Х2+Х3, … +Хn, Х0+Х(1+…n). This can get unwieldy when trying to save space in the data storage.

The difference between incremental and differential backups is easy enough to understand. In fact, it basically lies in one word. Take a look:

• incremental reserve copying processes files that have been changed or created since the latest backup;
• differential reserve copying processes files that have been changed or created since the latest full backup.

What other types of reserve copying exist?

Delta copying is considered to be a variation of differential backups. This server backup method only overwrites the changes in the files, not entire files themselves. However, it only works for files that have been changed, not created, which is why new files are copied fully.

This method creates copies quickly, saves storage space and produces less redundant data compared to incremental or differential backups. One would think that everyone would be using delta copying; however, that is not the case. These kinds of backups can only be created using specialized software. Recovery takes a long time as well: the data reconstructed from a mosaic of changed pieces. Nonetheless, this method is useful for guaranteeing uninterrupted data safety in cases when the backup is created as soon as the file is created or changed, or when there is a limit on bandwidth used to transfer copies to the remote storage.

The binary patch method works similarly to delta copying. It copies parts of changed files but uses a different comparison base (bits as opposed to delta’s blocks). Keep in mind that both methods are used along with differential and incremental copying, not on their own.

Mirroring technology is sometimes considered to be a type of reserve copying. This method is used on the hardware level in RAID1 or when creating mirror websites. In reality, mirroring is simple file copying with no archiving or systematization.

Over the last 12-15 years, reserve copying technology has undergone many critical changes that have led to the reconsidering of the effectiveness of old methods and the creation of new ones. For example, the implementation of snapshots of file systems that can be used to create a reserve copy allows for quick and convenient reserve copying in the cloud without stopping the virtual machine. When used in the cloud, snapshots are great for saving storage space, as they take up no space on the client’s drive.

How do I choose a reserve copying system?

If you want to pick a reserve copying system on your own, experts recommend that you keep these 4 universal criteria in mind:

• efficient resource usage: the system must work as autonomously as possible with a minimal load on the system’s resources and create the copies in as little time as possible;
• recovery speed: the software must recover your data from a reserve copy as fast as possible to keep your business processes running. Give preference to software that allows you to work with copies directly;
• data safety: backup software must provide a high level of security, both by cryptographic and hardware-based measures (secure channels in the storage, protection during the process of copying, the ability to recover an interrupted session);
• flexibility: the software must be usable with all types of data, as it is impossible to foresee what kind of data will end up critically important, as well as give you a choice of backup methods while performing well with all of them.

Modern software used by professional system administrators always meets these criteria. Besides, experts can help you pick the optimal backup method for each particular case. This helps avoid situations when a company loses all of its tools because of a poorly set up reserve copying systems. Learn more about how our engineers prevented this risk for a client who has experienced a full system crash before in our cloud migration & reserve copying case study.

Did you like the article?