![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
Entry tags:
Backup intervals
When I realize that I need to restore files from backup it tends to be either quite soon after they were deleted or a long time afterward. So, in organizing backups, I tend to adopt a somewhat logarithmic spacing: I may have a few backups from the past couple of months but also some even from past years. While I back up regularly they overwrite old backups so as to appromixately maintain such spacing.
I recently learned that my approach may be unusual. For instance, my impression is that one important set of backups at work goes no further back than three months. For me, the approach of equally spaced backups cycling through the available capacity is decidedly suboptimal: I want the resulting progression to look rather more geometric than arithmetic even if that means fewer backups in total from retaining some larger increments.
One hopes that this question has actually been studied and modeled properly based on the modification time of files that users need retrieving from backups. If so, I wonder what the findings were.
I recently learned that my approach may be unusual. For instance, my impression is that one important set of backups at work goes no further back than three months. For me, the approach of equally spaced backups cycling through the available capacity is decidedly suboptimal: I want the resulting progression to look rather more geometric than arithmetic even if that means fewer backups in total from retaining some larger increments.
One hopes that this question has actually been studied and modeled properly based on the modification time of files that users need retrieving from backups. If so, I wonder what the findings were.
no subject
These snapshots are hourly for 48 hours, daily for a week, and weekly for a month. (There are further snapshots, but not as readily accessible).
I also have no data but I agree with your impression that this better fits the pattern; you either realise you have made a mistake immediately and want a very recent backup, or you realise it much later and want the backups to go back a long way.
Increasingly, however, I find another factor is that most of what I do is in DVCSes, though, which both obviates the need for backups (provided the VCS is properly D) and guards well both against mistakes spotted immediately and ones spotted years later.
Where DVCS lets me down, I think, is that I want to be able to D random intermediate states even if they don't make any sense, without doing a lot of "git rebase -i" afterwards to make the history only consist of working versions.
no subject