Newsgroups: comp.periphs.scsi
Path: lth.se!sunic!mcsun!uunet!maxed!root
From: root@maxed (0000-Admin(0000))
Subject: Re: AMI EISA SCSI Cached Controller - Experience?
Organization: American Micro Group, Inc., Ft. Lee NJ
Date: Sun, 23 Aug 1992 17:28:52 GMT
Message-ID: <BtG5w4.AEH@maxed.amg.com>
X-Newsreader: Tin 1.1 PL5
References: <713809508snx@echidna.adept.oz.au>
Sender: ed@maxed.amg.com (Ed Whittemore)
Lines: 30

Stephen Hodgman (smh@echidna.adept.oz.au) wrote:
: Hello there,
: 
: We are considering using one of thes ona SCO Unix system here.
: Is there anyone in the U.S.A who has experience with this under SCO ?
: 
: OR
: any general experience with SCSI and SCO.  We have found that occasionally it
: cannot reat a block and the only recours is to rebuild the O/S rather than map
: out the sector.  Any comments  or help ?
: 
: Thanks
: Stephen Hodgman                 E-mail: smh@hakea.adept.oz.au
: Adept Software                   Phone: +61 6 285 3460
: Australia                          Fax: +61 6 285 3459

Our experience is that cacheing controllers under Unix are a waste
of money, and time, since benchmarks under load for multiple
users with simultaneous heavy disk use show no improvement over
non-cacheing Adaptec or BusTek models. Some discussion has it that the
added step of writing thru Unix' software buffers is the reason.

I'd like to hear from anyone as to their experiences in ths regard.
We've tried the DPT and AMI models with 4 and 16 MB or so of
cache under ISC 3.0, with small, and larger values for NBUF and NHBUF
and found this to be true.

-- 
Ed Whittemore   uunet!maxed!ed   ed@maxed.amg.com
American Micro Group, Inc.   201-944-3293
