Thursday 27 November 2014

Best Practices: Data Backup Strategy

While I was working with as Pre-sales Consultant with a Backup Software Company, the most common question I use to come across was "How should I backup my Data?" or "How frequent I should Backup my data?" or "what Strategy I should use to backup my Data?"

Many companies these days define a data backup policies with the retention period or type of backup during a certain period. Many of the backup strategy are defined by users say "Tower of Hanoi" or "Grand-Father-Son (GFS)", along with options like One time Full or Immediate Incremental backup, which are usually manual or with defined schedule.
However, I feel custom options, are for those, who really want to extract & use this software as efficient, as it is, to get complete ROI out of this application. Their are many dependencies or I can say "check list" before defining the a True Backup Plan. Space & Retention period plays the main role while defining the backup plan. While considering these parameters, we should also not forget what your software can do for you, like its Compression rate, around 55%, which I feel is the maximum I have seen as compare to other players in market. Apart from these, we should consider, am I looking for Data availability or Business Continuity. I should also consider RTO (Recovery Time Objectives) for the backup, the more number of backups we will have, the more time it will take to recover, which is standard for any application. These days some Backup Recovery products provide speed of around 1.3GB per minute. 
To clarify more about it, lets think of a scenario, where I have to 200GB of Data to be backed up in 2TB of Space across SAN. Now as per company policy, i have retention period of 6 months.  I am considering data is coming from a single server and we are doing the Image backup, not the files or folder. Now, the moment I will be doing 1 full backup in a month, with 4 differentials in week & daily incremental, I dont think we can achieve our target for keeping backup for 6 months, as usually differential backups are half the size of full backups. This doesn't mean option for GFS is bad, its just its the right choice in every situation. Backups are critical and part of every DataCenter compliance. They should be well defined, tested and implemented, considering the needs of the user. Backup Software is just an application, which can make wonders, but how to make wonders, its up to us. 
Coming back to the scenario, if the user would have selected the Custom Backups, with 1 full backup & daily incremental backup, it would have been enough. It will also provide me Business continuity as I will have less number of backups to be recovered and I can extract any file as well, hence have my data availability as well. 
Please Note Though sometime speed of backup also matter to us, but I am ignoring this fact since currently we are discussing the backup strategy, as speed is already pretty good as observed around 1.3GB per minute. However, we should consider the speed of backup application & connected storage speed of data transfer, when we have multiple backups running simultaneously. During that stage, we have to consider options like bandwidth allocation & de-duplication as well. 
Other Best Practices, I can think of, are as following (Please note, this is Purely on my Experience basis)
1. Do not give the passwords, wrong password attempts may corrupt the backup or can interrupt the backup operation when running a schedule. 
2. If number of machine is above 20, its better to create individual plan for group of machines. You may also consider creating different folder for backup in same location. Anyhow the backup plan waits for completing the previous plan, so like I said, its up to you, how you want your software to work.
3. Tapes are Pre-Historical :) (slow), These days backup application are New Generation (very fast), I dont prefer Tape if duration of backup recovery matters to you. Even though backup speed can reach up to 2GB per minute, but since target speed is low, data backup will be slow as well. However, they are still required but as offsite copy, should not be used as primary. 
4. Dont backup single machine on a de-duplication enabled location, since it will check for data redundancy in same location and will take time to complete backup.
5. Make sure you are using option to validate the "Full Archive" 
6. I always suggest to take backup on two different locations, if possible with no financial constraints to invest for data availability under Disaster Level 5. 
7. I suggest to take full partitions backups rather then filer or folder, since even though backup is of drive, we can still do Files & Folder recovery. Also, its much faster as does backup sector by sector and auto include any new file created in same folder location, which infact does not happen if we do files & folder. 
8. Drill for Test recovery should be done oftenly, atleast once in 3 months. 
9. Bootable CDs should be kept ready with latest version of kernel. Problems usually dont knock door before coming. We should be armed to fight against them. 
10. Notification plays important role. Make sure you are getting notification via email or SNMP. If its SNMP, make sure your Trap catcher is listening alerts. 
Feedback, Questions will be Appreciated. 
Note: This article is re-posted, earlier it was posted on vendor site. Its still a Hot thread in their community. FYI https://forum.acronis.com/forum/17555 

Any more questions? please write back or comment here. There are more things to share.. 

Request you to join my group on Facebook & LinkedIN with name "DataCenterPro" to get regular updates. I am also available on Tweeter as @_anubhavjain. Shortly I am going to launch my own YouTube channel for free training videos on different technologies as well. 

Happy Learning!!

No comments:

Post a Comment