Path: utzoo!mnetor!uunet!husc6!cmcl2!brl-adm!umd5!purdue!i.cc.purdue.edu!j.cc.purdue.edu!pur-ee!iuvax!isrnix!greg
From: greg@isrnix.UUCP (Gregory Travis)
Newsgroups: comp.sys.amiga
Subject: Hard Drive Questions (XEBEC) (longish)
Message-ID: <883@isrnix.UUCP>
Date: 23 Feb 88 16:08:14 GMT
Organization: Inst. for Social Research (Indiana University, 47405)
Lines: 61
Keywords: DMA A2090 Filesystem XEBEC

<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
<> This is a reposting, the original didn't seem to make it out <>
<> of Bloomington!  Apologies to anyone who's seen it already.  <>
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>

I have a number of questions regarding my hard disk.  The disk is
a Xebec (9776 I think, I'm at work and the 'puter is home).  Anyways, it's
a 20 meg drive which I believe is manufactured by NEC (I opened it up
once, just to peek).  I bought it way back when hard drives for A1000's
were rare and it had the best published real transfer rates.
  1) The drive is rated (according to the Xebec manual) at over 150K
     bytes/second.  Actual read transfer rate, going flat out, to my Amiga
     seems to be around 25-30K, which stinks.  I realize that the slow
     seek time (85ms) may account for some of this.  Will the
     new hard disk filesystem help out at all?  Will it even work with the
     Xebec software?
  2) I have noticed that when WRITING to the disk (say from Emacs
     or a utility program) the activity light on the disk is
     only on about 10-15% of the time spent writing - which indicates to me
     that the software is not able to keep up with the turtle-powered
     drive!  What could be going on?  I 'addbuffer' about 100 buffers for
     the drive (which work out of my FAST memory!  I experimented with
     changing the BufMemType in the mounttable from 3 to 5 and it worked.
     The Xebec manuals say nothing about this.  Needless to say the gained
     chip memory was welcome).  However, the above behavior (sleepy writes)
     happens with any number of buffers.  This behavior (idle drive even
     with outstanding I/O) also happens during reads, but then it is less
     noticeable.  Is the activity light a sign of seek activity?  Could
     the drive be writing to a single track over and over again (since the
     interleave is 5:1 (yuck))?  Even so, a full track write should take
     less than 1/4 second!  This is based upon a 2400 RPM rate and 10
     revolutions to write the whole track (5 revolutions of interleave and
     1 revolution each time just to account for any missed revolutions).
  2) Is there any possibility that the drive will work with a
     A2090 DMA controller?  I may buy a 2000 some day and I would like to
     keep the Xebec.  It is a "modified SCSI" (whatever "modified" means).
  3) Does anyone know what I have to send, and to whom (i.e. to SCSI.DEVICE
     or DH0:), to park the drive?  I would like to write a program which
     sits in the background and parks the drive every 10 minutes or so.
  4) Where can I get information on the controller itself?  I would like to
     know things like where in the address space the controller appears,
     what the registers are and what bits do what (if this sounds wrong,
     it's because I write PDP-11 UNIBUS drivers all the time and have never
     written a 68000 driver).

Primarily, I am interested in improvements to the system software so that
any I/O related waits are due to waiting for the DRIVE to do
its work.  I can't imagine why the drive would ever sit idle waiting for
the CPU to tell it what to do.  Even if the filesystem is not optimized
for hard disks, shouldn't the disk be busy seeking all over the place?

System is an A1000, 2.5 megabytes of ram, 68010 processor.

Hi Karl!

I remain,
-- 
 Gregory R. Travis
 Institute for Social Research - Indiana University - Bloomington, IN 47405
 ihnp4!inuxc!isrnix!greg
 {pur-ee,kangaro,iuvax}!isrnix!greg