Date: Tue, 24 Nov 1998 08:37:27 -0800 (PST) From: David Wolfskill <dhw@whistle.com> To: freebsd-isp@FreeBSD.ORG Subject: RE: Tape Library 2 Message-ID: <199811241637.IAA08141@pau-amma.whistle.com> In-Reply-To: <813A3F0E2D02D211884900A0C966731E4190F1@exchsrvr.inficad.com>
next in thread | previous in thread | raw e-mail | index | archive | help
>From: Joey Miller <joeym@inficad.com> >Date: Mon, 23 Nov 1998 10:26:15 -0700 >It depends on your needs. Standard answer #0.... :-) >I just finished building the backup routine >here, and the main stumbling block I ran into was tape speed and >capacity. And this is with new 8mm AIT (25gb native, 50gb theoretical >compressed). The drive only writes tapes at about 3MB/s (6MB/s >compressed), and takes almost an entire day just to dump 4 days of >backups to tape (about 50-60 gig). That seems a little odd to me -- as in taking rather more time than I would expect. I'm using amanda; the tape hardware is a DLT 4000 (1.5 MB/sec raw). I have things configured so that I'm doing a full dump of each filesystem on the Engineering net every other day (and a level 1 on the off-days). At the moment, I'm only dealing with UNIX boxen (fortunately); the amount of data being backed up is about 42 GB (going by the last couple of days' reports, using the size before I do any compression). The total amount of time (for the combination of the 2 days) is: Dump time: 8:45 (hr:min) Output Size: 22356 MB Original Size: 43800 MB Avg. Dump Rate: 230 KB/sec Avg. Tape Write Rate: 1250 KB/sec Now, one of the key things amanda does to accomplish this is -- whenever possible -- write the "dump" images to disk, then (asynchronously) spool them off of disk and onto tape. This permits multiple dumps to occur at the same time; it also makes it easier to keep the tape streaming. (Reason the tape is as slow as it is for me, I believe, is that I have yet to figure out how to force the tape drive to stop trying to re-compress already-compressed data. Haven't had the time to order the manual from Quantum & delve into the SCSI internals.) >Almost all of my experience with autoloaders and DAT in general have >been bad, btw. And I use DDS drives at home, with a 4-cartridge autoloader; no problems. (That's on a Sun SPARCstation 5, using a home-grown Perl script. amanda wants to write beginning at the beginning of a tape for each execution, and I don't have enough media for that. As it is, 4 cartridges is just about enough to handle 4 weeks' worth of full backups each Sunday, with incrementals back to the most recent full on each of the other days of the week. Using amanda, with a similar level of coverage, 4 cartridges would only be adequate for 4 days, rather than 4 weeks.) One thing that I find slightly annoying, but (again) I haven't had the time to work on it (enough): while the Solaris 2 "mt status" command reports the current file where the tape head is, the FreeBSD command does not. My home-grown Perl script uses "mt status" rather liberally to track the current actual position of the tape, and reports the information. It is my fantasy that this would make recovery rather less painful than it might otherwise be. :-} david -- David Wolfskill UNIX System Administrator dhw@whistle.com voice: (650) 577-7158 pager: (650) 371-4621 To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-isp" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199811241637.IAA08141>