Full, incremental, differential… Which stock to choose?
The purpose of most backups is to create a copy of data so that a particular file or program can be restored after data loss, corruption, or deletion, or even after a disaster. Backup is not a goal, but a means to achieve the goal of data protection. In theory, the principle of providing a backup copy is simple. But implementing an effective set of backup-related operations is relatively complex.
Software solutions dedicated to backup – or backup in English – reduce this complexity. However, data backup is only one part of a larger incident protection plan. Thus, a backup solution will only help revive operations after an incident due to careful design and testing, which is just as important.
Backup programs have historically offered many ways to do this. The most common methods are full redundancy, incremental redundancy and differential redundancy. Synthetic full backups and mirroring are also available.
Choosing the right strategy depends on several factors. For example, if you’re backing up to the cloud, incremental backups are usually better because they use fewer resources. You can start with a full backup in the cloud and then move to incremental backups. Mirror backup, on the other hand, is a more appropriate approach to in-place backup and is limited by the number of available disks.
What is a full backup?
A full backup is a basic operation. As the name suggests, it involves copying all data to a storage device such as disk or tape. The main advantage of performing a full backup with each backup is that an exact copy of all data is available from a single medium. This results in a minimum time to recover data, a metric known as Recovery Time Objective (RTO).
But there are downsides. A full backup is the longest process; sometimes it takes 10 times longer to complete than other processes. In addition, more storage space is required.
Therefore, full backups are usually performed only periodically. Data centers with small amounts of data (or critical applications) may choose to run a full backup daily or more frequently. Generally, backup operations use a full backup along with incremental or differential backups.
what is a reserve iincremental ?
An incremental backup only copies data that has changed since the previous backup. Generally, it works by comparing the timestamp of files in production with the last backup.
Since an incremental backup only copies new data since the last backup, the process is much faster than a full backup, so businesses can run it as often as they want. In addition, an incremental backup takes up less space than a full backup, meaning less media is used.
What is backup? differential?
A differential backup is similar to the first incremental backup: it copies all data that has changed since the last full backup. And it does this every time, regardless of the differential backups done at that time.
So, copying all the data that has changed since the beginning every time, a differential backup has the disadvantage of taking longer to complete than an incremental backup, but not as long as a full backup. Its size will also be larger than incremental stock.
What are the possible combinations?
Let’s summarize. As shown in the table below, each backup process works differently. A company should run a full backup at least once. The best practice we’ve come across is to perform a full backup every week, usually over the weekend, so that transferring the copied data over the network doesn’t penalize user traffic.
The following backups will ideally be daily, performed overnight. Typically, this will be either an incremental backup or a differential backup.
In this example, the first partial backup is performed on Monday evening. Whether differential or incremental, it will contain the same information. Starting with the third backup operation on Tuesday evening, data backed up incrementally contains only changes made since the last incremental backup, i.e. the day before. Although the third differential backup remembers all changes since the last full backup, the changes from the previous weekend.
These three backup types are used to customize the most optimal strategy for comprehensive data protection. Each alternative involves tradeoffs between performance, data protection levels, total amount of data stored, and cost. The table below gives a numerical example of the area required by the backed up data under three possible scenarios, showing which backups are needed to restore the activity. These calculations are based on 22 business days per month and one month of data retention.
As mentioned above, performing a daily full backup requires the most space and takes the most time. However, more common copies of the data are available and less media is required to perform the recovery operation. Therefore, implementing this backup policy is more disaster tolerant and ensures recovery in minimum time.
Another solution is to perform a weekly full backup along with daily incremental backups, which both achieves the shortest backup time during business days and uses the least amount of storage space. .
However, there are fewer copies of the data available and the recovery time is the longest, as a business may need to use up to six media sets to recover the necessary data. If data is requested from data backed up on Wednesday, Sunday’s full backup plus incremental media sets on Monday, Tuesday, and Wednesday are required. This can significantly increase recovery time and requires each Media Set to work properly; failure of one backup set can affect the entire recovery.
Performing a weekly full backup plus a daily differential backup gives intermediate results. This is because recovery requires more backup media sets than a full-day policy, but less than an incremental-day policy. Also, recovery time is less than daily incremental backups and longer than daily full backups. Restoring data from a given day requires a maximum of two media sets, which reduces the time required for recovery and the risk of problems with an unreadable backup set.
What is a mirror backup?
A mirrored backup – or mirror – can be compared to a directly usable copy of the source data. According to a manager at Nakivo, a backup solutions vendor, “All the different backed up files are stored separately, just like at the source. Therefore, this type of backup is called a mirror of the source data. »
One of the advantages of mirroring is the very fast data recovery time. Accessing various saved files is also easy.
However, one major drawback is the amount of storage space required. With this additional storage, businesses must be wary of increased costs and maintenance requirements. Also, if a problem occurs in the source database such as corruption or deletion, the mirror backup also suffers the same fate. Therefore, it is best not to rely solely on mirroring to protect your data.
RAID 1 is a type of mirroring known at the disk level. This process replicates data across two or more disks. Disk mirroring is an attractive option for data that requires high availability due to its fast recovery time. It is also useful for disaster recovery due to its immediate replacement capability. Disk mirroring requires at least two physical disks. If a hard drive fails, the company can use a mirror copy. Although disk mirroring provides comprehensive data protection, it requires a large storage capacity.
Which strategy is used in which situations?
For businesses with small databases, performing a daily full backup provides a high level of protection without incurring additional storage space costs.
Larger organizations or those with larger volumes of data or servers run a weekly full backup with daily incremental or differential backups as the best option. Using differential backups provides a higher level of data protection with a faster recovery time in most scenarios, with a slight increase in memory consumption.
But the strategy does not end there. In general, you’ll want to follow a so-called “3-2-1” best practice, which is to keep three copies of the data on two different media, with one extra copy.
In this regard, you can use a synthetic full backup, which is a more accurate copy of the backup itself. A synthetic full backup simply rebuilds the full backup image using any required incremental or differential backups. This synthetic full backup can then be stored on tape for off-site storage with the benefit of reduced recovery time as all data is consolidated.
Finally, remember that the purpose of a backup is to allow data to be restored later. So it is important to test your backups. Without periodic testing, it is not possible to guarantee that the data protection objective is achieved.
