Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!ucbvax!agate!shelby!portia!forel!karish From: karish@forel.stanford.edu (Chuck Karish) Newsgroups: comp.sys.ibm.pc.rt Subject: Re: Determining Density and Cluster Size for Tape Backup Message-ID: <5318@portia.Stanford.EDU> Date: 14 Sep 89 15:23:33 GMT References: <17911@ut-emx.UUCP> <186@bally.Bally.COM> Sender: USENET News System Reply-To: karish@forel.stanford.edu (Chuck Karish) Organization: Mindcraft, Inc. Lines: 44 In article <186@bally.Bally.COM> pete@bally.Bally.COM (Pete Gregory) wrote: >In article <17911@ut-emx.UUCP>, jjr@ut-emx.UUCP writes: >> I use "cat list | backup -i -f$DEVICE -d$DENSITY -C$CLUSTER" >> to backup files to a streaming tape cartridge. >> What are the appropriate values for $DEVICE, $DENSITY, and $CLUSTER? >> I have been using $DEVICE = /dev/rmt8, $DENSITY = 700 bytes per inch >> (default value), and $CLUSTER = 2000 blocks. /dev/rmt8 pre-tensions the tape, doesn't it? You probably don't need this if you re-use your tapes every few weeks. I usually use /dev/rmt0. >Why not use cpio? If your system gets trashed, you can re-install it from the standalone diskette if it's a backup archive. The standalone system doesn't include cpio. Read the manual for limitations. Beyond that, the choice of archive format ia a matter of personal preference, usually based on the user interface. IBM people tend to use backup for everything, BSD partisans use tar, USG users favor cpio. >We have had great success using HUGE buffers, huge enough >that we REALLY stream the drive (it never stops). Different utilities have different limitations on the maximum buffer size. Tar re-sets it to 800 blocks when I ask for more. The documentation is unclear, however, as to whether those are 512-byte blocks or 10Kb blocks, the default blocking factor for tar. I'll repeat my call for a generic tape-handling program that understands clusters. The manuals don't give enough information to enable me to write one myself. The worst limitation now is that dd doesn't understand clusters, so I can't use the drive efficiently to do a backup over the network with a command like tar cfb - 1 directories | rsh remotehost dd obs=20 > /dev/rmt0 What's needed is a program that would fit into the pipeline after dd and do the block/cluster management, like the `tcio' program that HP provides for their drives. Chuck Karish karish@mindcraft.com (415) 493-9000 karish@forel.stanford.edu