Veritas-bu

[Veritas-bu] Backup Schedules

2001-01-02 12:20:38
Subject: [Veritas-bu] Backup Schedules
From: Bob Bakh bbakh AT veritas DOT com
Date: Tue, 2 Jan 2001 09:20:38 -0800
Since no one else commented I will.

John this is good in theory, however this only would work on a file server.
If these servers work with a Database your method would not work so well,
since a Database touches all it's files and opens them all and changes
archive bits every time, you would essentially be doing a full every night
any way, unless your using BLIB.  Also you leave yourself in a bad place in
the case of a restore.  Imagine your system crashes 35 weeks and 4 days
after the last full you would need 40 tape mounts to be sure you had all
your data back.

In the other scenario, full once a month and cum weekly and diff daily, this
number could be as high as 9 tapes.  I'm a big fan of Full weekly and Cum
daily. this would allow for a two tape restore.

Just my opinion.

Thanks,

Bob

-----Original Message-----
From: John_Wang AT enron DOT net [mailto:John_Wang AT enron DOT net]
Sent: Wednesday, December 20, 2000 4:30 PM
To: veritas-bu AT mailman.eng.auburn DOT edu
Subject: [Veritas-bu] Backup Schedules




Hello

Seems to me that the general consensus is to do schedules of daily
differentials, weekly cummulatives and monthly full backups with retention
periods such that the differential expires before the cummulatives and the
cummulatives before the full backups.   How about the following ideas?

Establishing a baseline full copy with infinite retention.  Duplicating the
baseline.   Running daily differentials and weekly cummulatives with
retention
periods where the differentials expire before the cummulatives.   Whenever
major
application upgrades or other work has been done, a new baseline can be cut
and
the old one expired.   Standalone drives local to the site can be used for
the
baseline copies thereby allowing remote over the WAN operation where
bandwidth
is limited. (Hey, I was asked to try and back up London from here and it
would
take 28.5 days for a full backup so I need some optimization strategies).

Since in another thread, it has been alluded that incrementals are only
versus
backups in the same class as the incrementals.   If this is the case, two
classes could be established where instead of daily differentials,
differentials
are every two days, alternating between the classes.   The two classes could
go
to different pools or even volume groups thereby protecting against single
media
loss since the differentials are still being run each day but are versus not
the
previous run in another class but the one prior in the same class and hence
each
file revision is likely to get backed to two different volume pool/group
hence
separate tapes within two days.   Could make for automatic offsite policy if
the
two volume groups are geographically separate but well linked (Hey, I've got
four computer centers in this city alone and two are connected by gigabit).
Can anyone confirm if incrementals are class specific (obviously I'll set up
an
experiment but if someone knows, all the better).


Regards,
John I Wang
Sr. Systems Engineer
Steverson Information Professionals

---
Enron Broadband Services
3 Allen Center 3AC872e
ph (713) 345-6863



_______________________________________________
Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu



<Prev in Thread] Current Thread [Next in Thread>
  • [Veritas-bu] Backup Schedules, Bob Bakh bbakh <=