home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.unix.admin
- Path: sparky!uunet!gatech!usenet.ins.cwru.edu!agate!spool.mu.edu!uwm.edu!linac!att!att!dptg!ulysses!allegra!princeton!siemens!aad
- From: aad@siemens.com (Anthony Datri)
- Subject: Re: Dumping a Gigabyte
- Message-ID: <C16Ft8.Dw1@siemens.com>
- Sender: news@siemens.com (NeTnEwS)
- Nntp-Posting-Host: lovecraft.siemens.com
- Organization: Siemens Corporate Research, Princeton (Plainsboro), NJ
- References: <1j55eiINNk3d@BSDI.COM> <id.DZQW._25@ferranti.com> <ericw.727296638@angelo>
- Date: Wed, 20 Jan 1993 23:55:55 GMT
- Lines: 39
-
- >A full backup every week takes, say, 8 tapes for the full, and 5 tapes for
- >the incrementals (8 FULL+(1 INC*5days)=13tapes. Spreading that through the
- >month makes the formual (8 FULL+(1 INC*20 days)=28.
-
- Well, not really. From the man page:
-
- 0-9 The "dump level." All files in the filesystem that have
- been modified since the last dump at a lower dump level
- are copied to the volume.
-
- The usual scheme is to run daily incrementals at the same arbitrary level
- every day. Thus, if one did a level 0 monthly, one'd do a level 1 (or 5,
- or whatever) daily. Restoration 20 days (months have 20 days?) after the
- full would involve the full dump and only the most recent incremental.
-
- Of course, doing those incrementals sucks tapes, and toward the end of the
- month they're likely going to rival the fulls in size anyway.
-
- Has anyone ever actually used the infamous "tower of hanoi" scheme, or is it
- just someone's cruel joke?
-
- >2) Like most large shops, we run 7 by 24. The only time when we have even a
- >hope of getting clean backups is between 11:00pm and 7:00 am.
-
- I've always marveled when people say that they go single-user for backups.
- When people run three-week-long jobs, it's tough enough to even schedule a
- fastboot
-
- >3) DATS. Well, that's one I hadn't thought of. On the other hand, my
- >personal experience with them is that they are slower than the exabytes
-
- The Archive Python has a peak write rate of 183k/s. On an SS2 with a huge
- user-mode blocking factor I got about 150 with dump; about half that with
- tar. The HP-supplied drives that I've used on HP's seem to be *much* faster,
- but I have no numbers.
-
- --
-
- ======================================================================8--<
-