On Mon, Nov 22, 2010 at 7:43 AM, Binand Sethumadhavan binand@gmail.com wrote:
2010/11/22 Kshitiz kshitij_kotak@hotmail.com:
I feel a full backup once in 400 days, with hourly incremental, daily incremental and weekly incremental backup being rotated in a way in which 1 full backup followed by 24 hourly incremental backups, interspersed with 7 daily incremental backups and 5 weekly incremental backups may be protection enough for my data. What is the you all's expert opinion?
I am assuming you meant 40 days, and not 400 days. :-)
Let us say a full backup takes 30 minutes to restore, each weekly incremental takes 5 minutes to apply and each daily incremental, 1 minute. Let us further say you have an outage 33 days after your previous full backup. Now, the time for you to restore is:
30 minutes to restore the last full backup 4 x 5 = 20 minutes to apply the 4 weekly incrementals since the last full backup 5 x 1 = 5 minutes to apply the 5 daily incrementals since the last weekly incremental
Total time taken to restore = 55 minutes ~= 1 hour.
Nice analysis.
I feel we can provide better insight if he lets us know what exactly he is trying to backup. Since he is using rsync, I assume he is backing up files at the OS level. Now, what application does he use, and how often to files get generated/modified at the OS level? That would decide the frequency of his full and incremental backups.
If we are taking about OLTP here, then logic says the data be stored in a database, not at the OS level. In such a case, the db will take care of mttr, etc. The 55 minutes of data loss would be taken care of by the redo logs (if Oracle).
I have never come across a situation where I had to run hourly incremental backups at the OS level.
Regards, NMK.