Fujitsu 2392 with Xylogics 7053

[Curt Freeland]

jes at math1.mth.msu.edu (Jesse) writes:

>I would appreciate any information and experience [good/bad] about Fujitsu
>2392 2GB disk drive with Xylogics 7053.  Comparison between Fujitsu 2392
>and other similar drives would be nice, too.  I'll also need the
>format.dat entry for 2392.  Thanks in advance.

We are running two of these drives on Xylogics 753 controllers.  We had a
few problems associated with being "on the leading edge" and all :-).

If your controller is a pre-May 1990 unit, you probably need new
microcode.  The old microcode did not understand the way that Fujitsu did
extended cylinder addressing.  This same problem has been discussed here
on the CDC 9772/9773 disk drives (858Meg and 1.3Gig drives in the 14" form
factor).  The new firmware works fine for both drive types.  I have run
the following controllers, and all seem happy with new firmware:

753-103 (firmware # 180-002-186 Rev 2.23)
753-107 (firmware # 180-002-271 Rev 2.3)
7053    (firmware # 180-002-271 Rev 2.3)

We use our 2392 drives as server, and client root/swap filesystems, as
well as user filesystems.  They seem to be very nice drives.  There are
three problems (?) I am aware of though.

1) My hands are bigger than the designers.  This became painfully obvious
   the first time I tried to plug in the SMD interface cables.  I prefer the
   way the 2382 was laid out (you could at least get to the cables...  2392
   drive is a bit of a pain).  At least they have "A" cable "in" and "Out"
   connectors, instead of the one connector like the 2382 that requires funky
   cables.

2) No IPI interface is currently available :-(.  I hope someone at Fujitsu
   is reading this, and takes care of this problem!

3) The drives do not work on a Gould NP1 system, as the drive does the
   extended addressing a little awkwardly.  Actually, we are running one
   drive on a Gould, but we only use the first 1023 cylinders.  If we go
   beyond this limit, the drive does a RTZ for every seek to a cyl>1023.  We
   have been told by people at Fujitsu that they are working on a new version
   of the logic board for the drive that would fix this, and negate the need
   for the special Xylogics firmware.

Example format.dat entries:

disk_type = "Fujitsu-M2392 Swallow" \
        : ctlr = XD7053 \
        : ncyl = 1914 : acyl = 2 : pcyl = 1916 : nhead = 21 : nsect = 81 \
        : rpm = 3600 : bpt = 50400 : bps = 610

partition = "Fujitsu-M2392 Swallow" \
        : disk = "Fujitsu-M2392 Swallow" : ctlr = XD7053 \
        : a = 0, 20412 : b = 13, 74844 : c = 0, 3255714 : g = 57, 3158757

To use the settings above, set the SW1 for 608 bytes/sector (Switches
2,5,6) closed.  The drive automagically adds two to the value set by the
switches to give you the 610 bytes/sector.  And yes, I know the number of
sectors is actually 83+runt in our setup.  Slip sectoring, and paranoia
(read past experience) were included in the values we used....

Other drives?  Rumour has it that Seagate (was Imprimis, was CDC) is
readying a 2.5Gig drive with SMD-E and IPI interfaces.  I have not seen
one yet, but an associate has.  I am anxiously awaiting its release...

Curt Freeland
Manager, Systems Engineering
Purdue University
Engineering COmputer Network
(curt at mischief.ecn.purdue.edu)
(317) 494-3715