Subject: Re: Making use of the CI bus as IP data link ...
To: Gunther Schadow <port-vax@netbsd.org>
From: Geoff Roberts <geoffrob@stmarks.pp.catholic.edu.au>
List: port-vax
Date: 05/29/2001 09:13:48
----- Original Message -----
From: "Gunther Schadow" <gunther@aurora.regenstrief.org>
To: <port-vax@netbsd.org>
Sent: Tuesday, May 29, 2001 6:30 AM
Subject: Making use of the CI bus as IP data link ...
Hi there,
> while I'm waiting for material to get on with my VAX project I
> have some time for musings. Don't want to keep anyone from doing
> good work though :-)
Original thought is always valuable.
> I found a blue coax cable in the trash at work. Looked like new.
> About 7 or 10 mm thick in a roll of 1 ft diameter. Probably about
> 5 ft long (all estimates.) It has an inward-threaded connector
> at each end. I can actually screw those on my CI bus connectors.
> From what I heard, I imagined a CI cable being a big pipe of
> at least 1 inch diameter,
No.
> was I wrong
Yes
> and I do have a CI cable?
Sure sounds like it. CI is a semi rigid coaxial, something like RG8 in
size.
Always blue in colour.
> I know I would need a star coupler to do anything useful with
> this, right?
Right. Unfortunately, the CI design is such that it needs the current
balancing provided by the
star coupler. Direct connecting Vax to Vax may well lead to physical
damage.
> However, I'm not eager to run an HCS as I don't really want to
> build a cluster around common storage. Instead, I was wondering
> about other uses of the CI bus. I heard it is pretty impressive in
> terms of throughput, so would be a waste if it ended up unused.
Roughly equivalent to 80mb full duplex, and it has redundant paths.
Pretty slick for the time, yes.
> And since a 100-TP Ethernet card does not exit for VAXen,
There is FDDI stuff around, but not sure what speed it runs at.
> I wonder
> whether one could use a CI bus as a data link layer for IP?
Interesting question, in a cluster it allows itself to be used as
Telnet/Rlogin to
a HSC or another Vax, so I can't see why not. It would be a matter of
device
drivers I would think.
> Not
> having any documentation about it I can't really say how, and not
> even having a working machine (yet) it certainly is all theoretical.
> But I wonder that this super CI link technology is not put to more
> interesting use than just sharing mass storage (and if built with
> RA80's one can hardly even speak of "mass" anymore :-).
I've often wondered the same thing, only problem is that it is Big Vax
to Big Vax or HSC only in
the hardware area, though ISTR hearing there was now a CI interface for
one of the desktop boxen (later not MV 3100 vintage)
released about a year or so back. But if you had a couple of CI
equipped vaxen I don't see why it couldn't be used as a fast
network link. Antonio or the other DECsperts here may be able to tell
you more.
> So, I could imagine using CI as a link layer for IP just like
> Ethernet. Anyone ever tried that or knows otherwise that it
> cannot work?
No, as I said, it certainly provides a type of remote link even in it's
native form under VMS.
> My CI interface has 4 connectors "A" and "B" for each
> input "->0" and output "0->". Will one connection to a star
> coupler use all four of those or only two?
All four or you get error msgs. It has full duplex ability, but 2
redundant paths.
So you need 2 cables, 1 for tx 1 for rx, and they are duplicated for
redundancy.
It's probably possible to train it to ignore the missing duplex set,
again it would be
a question of the driver I think.
> How absolutely
> certain is it that one cannot make a direct link between two
> machines without a star coupler? Like you can make a direct
> link between two 10/100-TP ethernet ports with a crossover
> cable? One would simply cross the input connector of one machine
> over to the output connector of another machine and vice versa.
Sounds ok in theory, unfortunately the currents in the cable are not
such that it can be done
without the star coupler. A star coupler would be fairly easy to get I
think, since this stuff is
as obsolete as the systems themselves. What's also attractive is that
it is just a tin box in a cabinet,
and could probably be bolted into the vax itself, and just feed your
other vaxen from there.
It's a passive device (consumes no power) so there is no cost penalty to
running one.
> Does one have access to the physical layer with the CI
> adapter cards or is one confined to the MSCP?
Don't know, I imagine that would be a driver function rather than a
device function.
> I could imagine
> that the MSCP implementation of the CI controller has all kinds
> of built-in autodetection features that would refuse a direct
> link between computers, or even any communication between
> computers on the data link and network layers.
This is not so, as it will certainly allow me to log on to my HSC via
the CI, and
to any other Vaxen on it for that matter, with the SET/HOST command
under VMS.
> If so, it will
> not work. If, however, there is a way to bypass the MSCP or
> circumvent any auto-sanity-enforements, one could either implement
> IP over raw CI data link or IP over MSCP.
Sounds like it should be possible, in principle at least. Docs will be
required to figure out how.
> insane thoughts?
Not really. Only downside I see is that you need the star coupler, and
it's strictly big vax to big vax, fine if
you have a couple, not much use if you don't.
Cheers
Geoff in Oz