Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!ucbvax!decwrl!sun-barr!apple!rick
From: rick@Apple.COM (Rick Auricchio)
Newsgroups: comp.unix.aux
Subject: mkfs interleave factors
Message-ID: <31426@apple.Apple.COM>
Date: 23 May 89 17:29:51 GMT
Reply-To: rick@apple.com (Rick Auricchio)
Distribution: usa
Organization: Apple Computer, Inc.
Lines: 27

Just a quick note on what to use for the mkfs parameters for "gap" and
"blocks/cyl".  Use 1 and 1.

My tests have verified the observation that things are markedly slowed down
by the bogus defaults of 7 and 400.  I'm now changing those defaults to 1 and 1
for a future (post-1.1) release.

Use of 1 & 1 works fine on the drives I have here; Quantums have partial and
full-track buffering, hence they work well at 1-1 interleave.  Seagates don't
have onboard buffering, but A/UX 1.1 makes the 1-1 physical interleave anyway
with the Vulcan I/O acceleration code.

A/UX 1.0 may not be optimal with 1 & 1; it's possible that some interleaving
of filesystem blocks makes sense, but I haven't tried it.  I think it's better
to make the file system "right" so that it isn't horrible after an upgrade.

BTW, mkfs divides the "gap" value by 2 to get 512-byte blocks.  That makes
a gap of 1-3 become 1; 4-5 become 2; 6-8 become 3, etc.  Even dumber.

In conclusion, just use 1 & 1 for "gap" and "blocks/cyl".  This even improves
floppy performance by >2x.
-- 
Rick Auricchio, Apple Computer Inc, 20525 Mariani Av MS 27AJ Cupertino CA 95014
sun!apple!rick   OR   rick@apple.COM     Malibu N4364D     (408) 974-4227
  People who think money can't buy happiness are shopping in the wrong places.
My opinion is my own. My employer? They use a windsock and a fire extinguisher.