Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!linus!philabs!cmcl2!seismo!cit-vax!speck
From: speck@cit-vax.Caltech.Edu (Don Speck)
Newsgroups: net.unix
Subject: Re: Are these the disks for me?
Message-ID: <466@cit-vax.Caltech.Edu>
Date: Mon, 12-May-86 01:28:42 EDT
Article-I.D.: cit-vax.466
Posted: Mon May 12 01:28:42 1986
Date-Received: Wed, 14-May-86 06:50:02 EDT
References: <299@brl-smoke.ARPA>
Organization: California Institute of Technology
Lines: 40
Summary: ALL new products have bugs

In article <299@brl-smoke.ARPA>, lacasse@rand-unix.arpa writes:
> I would be careful on the the Spectralogic, as I don't think they have sold
> many controllers and you will likely find Unix software support lower
> because of this.
> Emulex is the clear leader.  You might consider their SC-7003.  Its a new
> product, 8 drives, 3.0 MB/S SMD.  We have one on order ourselves.

Emulex has as little Unix support as you can get:  last I heard,
they have no in-house Unix machines, only VMS.	So they have no
way of detecting those bugs that show up on Unix and not VMS.

A year ago we bought a Fujitsu M2298K and Emulex SC7000.  There's an
option to map the drive as two DEC RM05's (wasting space), or unmapped.
VMS has to use the former, but we wanted the latter; since we were
probably the first non-VMS site to buy a 2298 from them, guess what -
the unmapped mode didn't work.	A couple days, new prom revision.

At the same time we bought a tape drive and TC13 controller.
I started loading the 4.2bsd distribution tape and find that
standalone copy could read the tape, but the generic kernel
couldn't.  Much debugging later, I find that the TC13 is upset
because the status buffer just happened to cross a 256-byte
boundary - someone forgot to carry into the high byte of the
address.  Within the week, new prom revision (Rev. D).

For a long time I observed that the clock on that machine lost
many minutes whenever we did dumps.  I figured my non-standard
'dump' was overloading the raw I/O system, which I knew spent
lots of time at spl6(), but recently I figured out that it was
the tape controller's fault.  Another set of new proms (Rev. F).

(Emulex was extremely cooperative once they knew that there was
a problem, and in other respects their stuff has been great).

My point is that the latest, greatest board is unlikely to have
been debugged on a Unix VAX yet (they aren't as common as VMS).
You are asking to be a guinea pig.  It may be worth it	- it was
for us - but you'd better be aware that you're getting into that.

Don Speck	speck@vlsi.caltech.edu	seismo!cit-vax!speck