Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!wuarchive!texbell!ssbn!bill
From: bill@ssbn.WLK.COM (Bill Kennedy)
Newsgroups: comp.unix.i386
Subject: Re: Archive Tapes
Summary: use -C blocksize and -O device
Message-ID: <1444@ssbn.WLK.COM>
Date: 4 May 90 12:36:28 GMT
References: <29490@cup.portal.com> <1990May2.113532.26951@msuinfo.cl.msu.edu> <299@zds-ux.UUCP>
Organization: W.L. Kennedy Jr. & Associates, Pipe Creek, TX
Lines: 48

In article <299@zds-ux.UUCP>, bjstaff@zds-ux.UUCP (Brad Staff) writes:
> In article <7804@dmshq.mn.org>, pnessutt@dmshq.mn.org (Bob Monio) writes:
> [ Some stuff deleted ]
[ more stuff deleted ... ]
> 
> Interactive has done this with 386/ix.  Here is what I get when I run the
> following command:  (Note that I am using the raw, not the block, floppy
> device.)
> 
> $ find . -print | sort | cpio -ovB > /dev/rdsk/f0q15dt
> Reached end of medium on output.
> If you want to go on, type device/file name when ready.

You can avoid having to retype the device name if you use the -O option,
$find . -print | sort | cpio -ovB -O /dev/rdsk/f0q15dt
And when it's time to change diskettes all you have to do is press return.
I've not tried it, but I suspect you could specify a full cylinder for the
blocking factor with the -C argument and maybe have fewer physical I/O
events.

[ ... ]
> new one.  If your cpio doesn't do this, complain to your vendor.  Of course,
> your floppy and/or tape drivers have to be well behaved as well if this is
> going to work.  If they aren't well behaved, complain to your vendor.

If you're going to tape I highly recommend the -C argument.  It lets you
specify the physical block size.  The ISC cpio inadvertantly adds a zero
for you (I say -C512000 he uses -C5120000) so watch out that you don't
require more buffer than you have memory.  If you do that it will go to
swap space and you lose a lot of time splashing back and forth.  On ssbn
(10Mb) I use -C 512000 -O /dev/tape and on inebriae (4Mb) I use -C 51200
-O /dev/tape.  When it's time to put in the next volume I just wait for the
tape to rewind, stick in the next cartridge and press return.

> I hardly ever use tar, so I can't offer any suggestions there.

Me neither, but I have a neighbor who does and he gets his tape to stay
streaming.  Even at -C 512000(0) I get six or so physical records (tape
stops, writes record mark and gap and proceeds) per I/O event.  GNU tar
seems to pour it out all at once and he gets one tape record per I/O event.
> -- 
> Brad Staff               |
> Zenith Data Systems      | "A government that can forbid certain
> 616-982-5791             |  psychoactive drugs can mandate others."
> bjstaff@zds-ux.zds.com   |	- Russell Turpin
-- 
Bill Kennedy  usenet      {texbell,att,cs.utexas.edu,sun!daver}!ssbn!bill
              internet    bill@ssbn.WLK.COM   or attmail!ssbn!bill