Talk:Backup rotation scheme

Towers of Hanoi backup
This article needs a detailed explanation of Towers of Hanoi backup scheme. The explanation given on 18:29, 12 September 2007 was taken from W. Curtis Preston's "Backup & Recovery", published in 2007.

The book's explanation is quite difficult to follow anyway. Needs a good diagram. Stumbles 13:46, 1 October 2007 (UTC)

I'd like a explination of the Incremented media method too. Logomachist (but not signed in) @ --24.115.80.11 04:54, 12 November 2007 (UTC)

When Tower of Hanoi is used for incremental backups, it works differently -- the tower is reversed. The purpose then is to reduce backup time and storage requirements, at the expense of convenience. Incremental Tower-of-Hanoi is most useful for online backups on 24/7 servers, and where the backups themselves are backed up through more traditional means.

A typical incremental Tower-of-Hanoi scheme is:

Day 1: Full (level 0)

Day 2,4,6,8,10,12,14,16,18,20,22,24,26,28,30: Level 5

Day 3,7,11,15,19,23,27,31: Level 4

Day 5,13,21,29: Level 3

Day 9,25: Level 2

Day 17: Level 1

Typically, the level 5 backups will be very quick and small, making a minimum of impact to a production system. The main trade-off is that files that are created and later deleted within a set period can no longer be recovered once all the incremental levels it was on have been overwritten. Thus it is most often combined with a permanent backup of the incremental files. 12.34.212.202 (talk) 20:12, 9 January 2009 (UTC) 

The text about duration of backup saved is misleading at this point: "A set of n tapes (or tapes sets) will allow backups for 2 n-1 days before the last set is recycled. So, three tapes will give four days worth of backups and on the fifth day Set C will be overwritten; four tapes will give eight days, and Set D is overwritten on the ninth day; five tapes will give 16 days, etc. Files can be restored from 1, 2, 4, 8, 16, ..., 2 n - 1 days ago." The text gives the impression that n tapes can provide 2^(n-1) days worth of backup on all days. Actually, on the day that the nth tape is overwritten, the n tapes provide only 2^(n-2) days worth of backup, as can be seen by inspecting the tables shown. Recommended new text: "A set of n tapes (or tapes sets) will allow backups for 2 n-1 days before the last set is recycled, with the depth of the backup falling to 2 n-2 days on the day the last set is recycled. So, three tapes will give four days worth of backups and on the fifth day Set C will be overwritten and two days of backups will be available; four tapes will give eight days, and Set D is overwritten on the ninth day leaving four days of backups available; five tapes will give 16 days, and Set E is overwritten on the 17th day leaving eight days of backup, etc. Files can be restored from 1, 2, 4, 8, 16, ..., 2 n - 1 days ago just before the end of the cycle, and 1, 2, 4, 8, 16, ..., 2 n-2 days ago on the day the cycle starts anew."

128.149.22.24 (talk) 22:07, 4 March 2011 (UTC)

GFS schedule
There is a lot of confusion about this subject. I find many references that explain the Grandfather-Father-Son rotation schedule as follows:

1 Tape (/set) used for Grandfather = Full Backup 4 Tapes for daily incremental backups Monday to Thursday 1 Tape used weekly on Fridays = Father = Differential backup

My references then explain that Full and incremental backups clears the archive bit and that differential backups does not do this. Surely this means that every weekly Differential backup will only contain data that was changed on the Friday and the rest of the weeks info will be lost when those son tapes are re-used again! Or is there another method other than the archive bit used to determine the backup status for differential backups.

If on the other hand weekly incremental backups are made and daily differentials as some references suggest it would mean that more than one father tape would be required for a months worth of backups (i.e. not six tapes but 4+4+1)

After reviewing several online references dating back to 2001, I felt that the simple description was appropriate. Pdinhofer (talk) 19:00, 3 February 2008 (UTC) — Preceding unsigned comment added by IO Device (talk • contribs)

Weighted random distribution
"The Weighted Random method has no practical advantage over a more systematic approach." - I would like to know why this is thought to be the case. I'm looking at implementing a disk based backup and it would seem to me that it would be more likely to provide the best coverage, but if I'm wrong more information would be useful. --Tim abell (talk) 17:03, 18 June 2016 (UTC)


 * Indeed; you are not wrong. The randomness risks being difficult to analyze, and not predictably systematic as say weekly/monthly/quarterly/half-yearly/annual backups. --Hyperforin (talk) 00:43, 19 June 2016 (UTC)

Look what you're doing! That's doing your own research. If it goes on Wikipedia it needs a source, Hyperforin --2A01:C22:34FA:3500:981D:F474:79DD:990A (talk) 04:02, 20 March 2024 (UTC)

Article is way too long and detailed
It would be perfectly fine to refer to literature for further information. The reader should just know what the lemma means, not necessarily be able to act upon it. 2A01:C22:34FA:3500:981D:F474:79DD:990A (talk) 03:59, 20 March 2024 (UTC)