Having a backed up copy of your data, allows having them recovered, in case of emergency. But when to create that copy? How often to perform a backup? Do you need to backup every day? In this article we’ll talk about backup strategies – FIFO scheme and GFS scheme.
Backup strategies – start from the end
By intuition many of us would want to store as many copies of data as possible. But the problem of shrinking available storage space occurs really quickly.
When it comes to determining the number of needed copies, or frequency of running a backup, the best way to know your situation, you need to answer the following questions:
- What is the amount of storage available for backup?
- What data do you want to protect? From how many devices? Single files or whole systems?
- What is the size of that data?
But also:
- Is the storage big enough?
Thus we need to find a compromise between “protecting all” and “being secured enough”. Imagine a situation where you need to recover data. For example, storage on users’ devices breaks down. Would you in that situation want to recover the full image of the disc? Does it need to be a copy from yesterday, or it’s enough that it’s from last month? Does your user store data on-premise or in the cloud? Maybe in the event of a failure of any station, you will only need an image of any machine, where the system and the basic set of applications will be installed, and the user will log into their accounts? Maybe it’s enough to protect the most important files from the user desktop? So that there is only a need to recover the disc image and some of the catalogs for users to continue their work?
Prepared for everything
Creating every copy takes time, consumes network and devices resources from where data is being backed up. So the good solution is to minimize the amount of copies made.
When you know what data you need to protect, and the scenarios that can happen, it would be easier for you to know what to do.
Example scenarios:
- Due to software malfunction, needed records in the database were deleted. You need to recover the database from before the malfunction. But you noticed it 5 days after it happened. So you would need to recover the copy from 6 days ago.
- User deleted a folder from the shared location, and other users need it. You want to recover the last version of those files.
- Mailbox server had an accident. You want to recover the whole system. It can be a copy from months ago, because you backup Exchange independently every day.
- You need to quickly prepare a workstation for a new employee. You want to recover an image of the device, freshly after the installation of the system, but to another device.
- Ransomware affected the storage on few devices. You want to recover systems and files of the users.
We can spend the whole day trying to imagine situations, simpler and more complicated, that can happen in the organization. Generally, it’s good to make copies of the most important data (like databases) every day, and those less important we can perform in larger time intervals (e.g. image of the server disc every week)
Automating Backup
In the previous blog posts we talked about backup types. Using full copies, incremental copies, and differential copies (where full copy is performed after a set interval of copies), using retention – regularly deleting old copies after they pass a certain “age”.
FIFO Scheme
The simplest of backup schemes is FIFO scheme – first in first out. It’s premise is to create incremental copies between the full ones. You only store X-amount of copies and the older ones will be deleted. But this scheme assumptions don’t cover how often to perform the copies.
When it comes to incremental copies, the number of copies made, it can be increased by a few to ensure the safety of data. So that, there is no chance that during creating the full copy, the oldest one gets deleted by an accident, and all incremental ones originated from that full copy, making them not able to be recovered.
Pros:
- Easy implementation
- Copies can be stored for a really long time with a accurate time stamp
Cons:
- The longer the retention period, the more storage space is required.
GFS Scheme
Grandfather – father – son scheme allows to manage the copies in the long period of time.
An administrator separately decides for how long to retain full copies, incremental copies, and differential copies. The GFS scheme uses those three types of backup in one chain. The Grandfather copy is a full copy performed once a month. Those copies will be retained for 12 months. At the end of every week the Father copy is created – differential one, which is retained for a period of one month. And every day the Son copies will be created – incremental ones.
The great thing about this scheme is that the maximum number of copies needed to recover is 1 full copy, 1 differential one, and 6 incremental copies. This leads to fast recovery times. Also you retain the old copies, as well as saving storage space.
Pros:
- Long retention times
- Saving the storage space
- Fast recovery times – worst case scenario you need 7 copies to be consolidated
Cons:
- Strict backup schedule
The textbook approach to the GFS scheme has a strict time schedule for performing copies and their retention times.
Today backup solutions providers create tools allowing you to more flexibly manage copies. Allowing you to create individual rules within GFS scheme applied to performing a backup, and retain copies.
Backup strategies – time to decide
Which scheme is better? There is no simple answer to that question. In some cases FIFO will be the more suited one, in other GFS. Backup strategies need to include key needs of your organisation when it comes to protecting your data.
For backups, you want to retain for only a couple of days the better choice will be FIFO scheme. This way you can save storage space. On the other hand, if you prioritize recovering the data from the exact point in time, more than storage space, you can use FIFO scheme with a longer retention time. The GFS scheme works better when you want to retain copies for long periods of time.