Great it really works!!

Rick

On Wed, 25 Apr 2001, Robert Elz wrote:

> Date: Wed, 25 Apr 2001 23:25:20 +0700
> From: Robert Elz <kre@munnari.OZ.AU>
> To: Riccardo.Veraldi@fi.infn.it
> Cc: Ray Phillips <r.phillips@mailbox.uq.edu.au>, port-alpha@netbsd.org
> Subject: Re: AlphaStation 255
>
>     Date:        Wed, 25 Apr 2001 17:07:56 +0200 (CEST)
>     From:        Riccardo.Veraldi@fi.infn.it
>     Message-ID:  <Pine.NEB.4.33.0104251705270.10640-100000@nikita.fi.infn.it>
>
>   | I installed a IBM DNES 18GB and when I installed the system the default
>   | block size and frag size of 80192 amd 1024 where not fine in my case.
>
> Something is wrong with your disklabel - perhaps you have defined tracks
> that are way too huge, or something odd like that.   The cyls/tracks/sectors
> numbers aren't used for anything much related to any modern drive,
> but they can make a huge difference to the way the filesystem works.
>
> Find numbers that multiply as close as possible to the size of the drive
> (but less or equal, not more), where sectors/track is of the general order
> of 64 (something between 32 and 80), and tracks/cylinder is of the order
> of 20 (10-24 is generally good).   Those numbers match the general layout
> style of the SMD drives from which all of this was originally designed.
>
> For what it is worth, "factor" is an incredibly useful program for
> calculating this stuff.  Eg: if you have a drive with 11733120 sectors,
> factor will produce 2 2 2 2 2 2 2 3 3 3 5 7 97 as the prime number factors.
> >From that, sects/track=64 (2^6) tracks/cyl=18 (2*3*3) cylinders=10185
> (3*5*7*97) would be a nice layout to try.  (64 * 18 * 10185 == 11733120)
>
> On an i386 you have to be concerned with the stupid bios limitations
> (so tend to have 63 sects/track, 15 heads (tracks/cyl) etc) - but on
> an alpha there is no suck stupidity to bother with, so just make the
> nicest numbers that fit.
>
> If factor produces just a few large numbers as factors, then try again
> with a number a few smaller (making it be a multiple of 128 is generally
> good where possible, or 64).
>
> kre
>
>