NetNews Usenet Archive 1992 #27/NN_1992_27.iso/spool/fj/maillis/apollo/2971

Text File | 1992-11-17 | 3.6 KB | 74 lines

Newsgroups: fj.mail-lists.apollo
Path: sparky!uunet!stanford.edu!agate!spool.mu.edu!sgiblab!nec-gw!nec-tyo!wnoc-tyo-news!etl.go.jp!daemon
From: [e-mail] (jt -- John Thompson)
Subject: re: Need Some Advice..
Message-ID: <[e-mail]>
Sender: [e-mail]
Organization: Electrotechnical Laboratory, Tsukuba Science City
Distribution: fj
Date: Tue, 17 Nov 1992 16:34:06 GMT
Return-Path: <[e-mail]>
Approved: [e-mail]
Lines: 60



> |> >If you want to make the assumption that IP packets are encapsulated within
> |> >DDS packets, then in order to make HP-UX reside on ATR, HP must then have
> |> >ported DDS to HP-UX on 400's and 700's... That must mean 700's have
> |> >node-ids. Somehow I doubt that.
> |>
> |> Actually, I understand that although ATR cards are available for 700's, they
> |> use a different protocol than standard Apollo DNxxxx* nodes, and a 9000/400
> |> series node is needed to translate protocols.  I assume at least the packet
> |> structure is different, but don't know all the details.  I hope to find out
> |> soon; if anyone has any more complete information ...
> |>
>
> It's my understanding that the 700 ATR cards talk TCP/IP only.

OK -- more complete (but still user) information:

The ATR card in the 700 series is similar to the ATR card for PCs.  It has its
own nodeID on the card, and responds to just enough DomainOS network commands
to get in your way, without helping you enough to make it worth the trouble.
For example, it will respond to lcnode, and if you catalog the nodeID, it will
even respond with a name.  Those of us who have scripts which do an lcnode and
execute a command for each name returned now have broken scripts.  In addition,
if you catalog the nodeID as //hpux_node, you can't NFS mount the file system
into the root area as //hpux_node.  The names conflict (unlike a Domain/OS
machine, where the node name and the file system are tightly linked).  The
workaround, of course, is to catalog the ATR card as //hpuxnode_atr.  Then you
can rewrite all the scripts to 'fpat -x //?*_atr' before operating.  <SIGH>

As far as protocols, the HP700 w/ ATR card handles TCP/IP only, except to have
enough smarts to pass the DDS network-requests around the ring (e.g. lcnode).
You cannot make a 700 series node act as a DDS gateway, although it acts as a
very nice IP gateway.  (BTW - a DN10000 makes a pretty nice router too, both
for DDS and IP networks.)

I believe the original posting was on NFS mounting the Domain super-root on
a Unix box, versus having each Domain node offer their file systems, and mount
each one.  The issue was speed, and the question was whether the protocol
translations were slower than the packet gating, since most sites have an
ethernet that accesses the ATR via one or two nodes.  Without going into
details, I have found it to be about twice as fast to have each Domain system
handle NFS access to its own root only, compared to having the gateway node
offer NFS access to the superroot.  I'm hoping to do more testing/eval as I
set up automounts, and will post again later w/ more fun data.

-- jt --
John Thompson
Senior Engineer / Sys-Admin
Honeywell, SSEC
Plymouth, MN  55441
[e-mail]

If you're not part of the solution, you're part of the precipitate.
*************************************************************************
* This e-mail correspondence is a work of fiction.  Any similarity      *
* between the views presented and actual views, personal or corporate,  *
* living, dead, or still to be conceived, is pure coincidence.          *
*************************************************************************