delorie.com/archives/browse.cgi   search  
Mail Archives: opendos/1997/09/16/17:41:46

Date: Wed, 17 Sep 1997 09:38:38 +1200
From: physmsa AT cantua DOT canterbury DOT ac DOT nz (Mr M S Aitchison)
Subject: SCSI disks, NWCACHE and fragmented disks (was Re: ClosedDOS???
To: opendos AT delorie DOT com
Message-id: <199709162138.JAA17917@cantua.canterbury.ac.nz>

-= ArkanoiD =- wrote:
> > btw NWCACHE slows down terribly on fragmented disks.

I'd like to find out more details, e.g. how much RAM was being used by
the cache, how big the disk is, how many buffers are defined in
config.sys, etc.  It is often said that you can take the buffers down
to something really low, e.g. 2, when you have a cache.  This isn't
exactly true. It still helps to keep the buffers at something like the
number of files in the largest directory in your path divided by 16.
This seems to be especially true if the amount of RAM the cache gets is
below a couple of Mb and/or you are wasting the cache by reading large
amounts of data that rarely is needed again.  I find that usually
NWCACHE does a very good job, better than most, but all caching
algorithms make assumptions about how likely certain blocks are to be
needed again soon. Having buffers set high (use up all of the remaining
HMA at least) ensures that one important area - the directory - is
pretty likely to stay in RAM.  If NWCACHE isn't working well for you
I'd love to know why not, so please let me know if increasing the
buffers to over 20 (preferrably 40) helps, and any other details of
your setup you can feed into an email message!

Christopher Croughton wrote:
> I haven't bothered with SCSI.  Why pay twice as much for no noticable
> improvement over EIDE?

SCSI disks are much better in some situations, for example Linux can
arrange several disk requests at the same time. Normally, DOS-like
operating systems cannot do this (I'm thinking hard how OpenDOS could
be made to; feel free to jump in with suggestions). Even so, there are
still some throughput advantages with SCSI in that EIDE's quoted
maximum throughput is often a long way from what you actually get,
whereas a good SCSI controller and disk can be expected to come close
to the maximum.  Be careful not to go by published specs, but real life
situations... I'd agree that, with (Open)DOS at the moment and typical
legacy applications, you *probably* will find the extra cost of SCSI is
better spent in RAM or CPU speed, but most Linux owners (whether or not
they are servers, know SCSI is a good investment.

-------------------------------------------------------------------------------
Mark Aitchison, Physics & Astronomy   \_  Phone : +64 3 3642-947 a.h. 3371-225
University of Canterbury,             </  Fax   : +64 3 3642-469  or  3642-999
Christchurch, New Zealand.           /)   E-mail: phys169 AT csc DOT canterbury DOT ac DOT nz
#include <disclaimer.std>           (/'   Callsign: ZL3TQE
-------------------------------------------------------------------------------


- Raw text -


  webmaster     delorie software   privacy  
  Copyright © 2019   by DJ Delorie     Updated Jul 2019