Return to BSD News archive
Newsgroups: comp.unix.bsd.bsdi.misc,comp.os.linux.misc
Path: euryale.cc.adfa.oz.au!newshost.carno.net.au!harbinger.cc.monash.edu.au!news.mira.net.au!news.vbc.net!vbcnet-west!garlic.com!uunet!in1.uu.net!news.new-york.net!news.spc.edu!spcunb.spc.edu!not-for-mail
From: Terry Kennedy <terry@spcunb.spc.edu>
Subject: Re: BSDi + AHA2944W + Eclipse RAID
Content-Type: text/plain; charset=US-ASCII
X-Newsreader: TIN [UNIX 1.3 unoff BETA release 961018]
X-Nntp-Posting-User: terry
Sender: news@spcuna.spc.edu (USENET News)
Content-Transfer-Encoding: 8bit
Organization: St. Peter's College, US
Lines: 75
Message-ID: <E17w2C.ME8@spcuna.spc.edu>
References: <55j0ci$8l8@gol1.gol.com> <55tuo6$r02@news1.anet-dfw.com> <56e1qh$4t8@oldman.steinkamm.com> <56glrn$c1k@olympus.nwnet.net> <E10H2w.BC0@spcuna.spc.edu> <56qp4d$os2@olympus.nwnet.net> <E13pK0.2rF@news.interactive.net> <57088k$3ir@olympus.nwnet.net>
Mime-Version: 1.0
X-Trace: 848574227/28745
X-Nntp-Posting-Host: spcunb.spc.edu
Date: Thu, 21 Nov 1996 11:03:48 GMT
Xref: euryale.cc.adfa.oz.au comp.unix.bsd.bsdi.misc:5320 comp.os.linux.misc:142988
In comp.unix.bsd.bsdi.misc Anthony Talltree <aad@nwnet.net> wrote:
> Others have told me explicitly that the Tomcat II is identical except for
> the second, currently useless CPU socket. What problems does the Tomcat II
> have?
Tomcat II boards are the same logic board as the Tomcat I except that the
second CPU socket is populated and there's a dual-processor arbiter chip (part
of the Intel HX chipset) installed.
Tyan has had problems with the cache subsystem. This comes from four differ-
ent areas:
1) people were trying to use pre-HX COAST modules which were electrically
incompatible
2) People were trying to cache over 64M without installing extra tag RAM
3) Some vendors were shipping non-Tyan cache modules
4) Tyan's vendor substituted different cache module parts which caused a
number of problems
In order to prevent all of the above, the motherboard has been redesigned
to have the cache soldered on (no more COAST module).
> Absolutely *NOBODY* has been able to tell me what the Tomcat III *IS*.
> Tyan's web pages describe it as identical to the Tomcat II.
See above. You've been asking in the wrong places, I guess. This has been
discussed at length in the Tyan newsgroup, for example. A Tomcat III is the
same as a rev 4 Tomcat I/II with soldered on cache. It comes in both single-
and dual-processor variants.
> Speed isn't my concern at this point. Something that will stay up for more
> than a day before silently wedging is my goal.
Lots of us have systems with apparently similar hardware that don't ex-
perience problems.
> Do you really think that cables and terminators would cause the entire
> system to wedge with no messages?
Yes, at least in older releases. Now you'll probably get the "bha0: oper-
ation time exceeds N seconds". Silent wedges are still possible, though,
depending on how badly you wedge the BusLogic microcode.
> I see such claims often. Could you back it up by stating exactly what
> you consider high-end cables & terminators, and where they can be purchased?
This is "common knowledge". If you want to conduct the research to convince
yourself, go ahead. Start with el-cheapo cables and cut one open. Note that
there are only 25 wires inside (if that many!), they aren't in twisted pairs,
and there's no overall shielding or it doesn't completely cover the conductors.
Now take an ohmmeter and check all the signal returns for continuity with each
other. You'll see they're all soldered together. Next, get a scope and look at
the ack line (as a good example). Look for ringing, over- and undershoot con-
ditions. Remember that Fast SCSI has 100ns pulses on the cable and that Ultra
SCSI has 50ns pulses. Now ask anyone with experience in transmission line en-
gineering what these cables will do to those pulses.
Good cables have individual twisted pairs for each signal, the cable is im-
pedance controlled, there's a good overall shield connected to the connector
shells, and so forth. Next, you want to use active terminators if you're
using single-ended SCSI - differential uses passive terminators specially
designed for differential mode.
You can either cut open cables to test for these things, or you can buy from
vendors that you know make reliable cables. I have experience with cables from
DEC and from an outfit called C.S. Electronics. I'm also told that Granite
Digital makes good cables, though I don't have any of those.
Once you have good cables, make sure the rest of your SCSI bus is configured
right - no stubs, terminators only at each end, TERMPWR configured properly,
etc.
Terry Kennedy Operations Manager, Academic Computing
terry@spcvxa.spc.edu St. Peter's College, Jersey City, NJ USA
+1 201 915 9381 (voice) +1 201 435-3662 (FAX)