Sun-Spots-Request@RICE.EDU (William LeFebvre) (10/15/87)
SUN-SPOTS DIGEST Thursday, 15 October 1987 Volume 5 : Issue 51
Today's Topics:
Re: Overloaded Ethernet
Re: WYSIWYG editor for the Sun
Re: Forking a shelltool from a shelltool
DNA product
Looking for beta-sites to test Lucid Common Lisp for the Sun 4
X PostScript Previewer
Sun-4 bcopy warning
Sun csh weirdness
varargs: bug or feature?
Interrupts clog up Sun
SYNC devices on zs or mti ports?
conversion from IBM/SCRIPT to TeX?
Printcap entry for HP-Laserjet?
X.25 using Suns?
Mandelbrot icon
----------------------------------------------------------------------
Date: Thu, 8 Oct 87 00:35:48 CDT
From: tundra!prime!pfh@sun.UUCP
Subject: Re: Overloaded Ethernet
>From: becker@cs.rochester.edu
>...
>Assuming that our ether is overloaded, we're trying to figure out the best
>way of segmenting our ethernet....
This is a lengthy response to Mr. Becker's question. I would like to hear
about administration at other large sites, too.
[[ A large portion of the original text of this message has been edited to
hold down its size. The original message is available in the archives as
"sun-spots/Ethernet.advice". It can be retrieved via anonymous FTP from
the host "titan.rice.edu" or via the archive server. For more information
about the archive server, send a mail message containing the word "help"
to the address "archive-server@rice.edu". --wnl ]]
SUMMARY
Set up a hierarchy of Ethernets using gateway machines. Keep the diskless
machines on very lightly loaded Ethernet trunks. If you have a lot of
Suns, put them in clusters and restrict NFS to machines within a cluster.
Don't export *and* import NFS filesystems on the same machine.
NARRATIVE
Administration of Sun networks is a bit of a black art. A large
population of diskless machines requires a reliable and lightly loaded
Ethernet in order to survive. On the other hand, it's worth it. If you
haven't had to use IBM VM/CMS and MVS to do your work, you can't fully
appreciate Suns.
When I was at Cray, our Sun population grew to several hundred model-50s,
with a server for every ten diskless machines. The building I was in has
two three-story wings and one four-story wing. We started out with one
Ethernet trunk down each corridor, for a total of ten trunks.
This configuration turned out to be a problem as the net grew. Some of
the administrative problems were:
- A mixture of Sun-2s and Sun-3s
- The desire to have Sun-2 servers serve only Sun-2 clients
- Rapid growth (a new string of machines every few weeks)
- NFS problems (many caused by cross-mounting filesystems)
- Heavily loaded Ethernets hurting performance and NFS reliability
The mix-and-match problems were worst on the three floors that had the
most machines. We bought DEC DELNIs (dumb Ethernet repeaters) and hooked
the three local trunks together; at the time, this included 40-50 diskless
machines. This was a disaster, of course. ND was sluggish and NFS
complained frequently of timeouts.
Eventually, by trial and error, we hit upon a distributed system, which
is a hierarchy of Ethernets:
- A company-wide Ethernet (Suns and other equipment)
- Several cluster Ethernets (Sun servers only)
- Many local Ethernets (diskless Suns and PCs only)
The figure below shows a small cluster of three file servers and what
we call a "communication server:".... [[ figure omitted. --wnl ]]
We cleared up many of our NFS problems by making the file servers "pure"
NFS servers, that is, they export their filesystems and do not import any.
The comm server and all the diskless machines are pure clients; they only
import NFS filesystems. By the way, this was quietly recommended to us by
Sun developers.
Furthermore, we restricted NFS and YP to a cluster. (No NFS mounts across
the company backbone.) Each diskless machine mounts all the filesystems
in its cluster. The NFS filesystem(s) from the diskless machine's ND
server are hard mounts. All the other NFS filesystems are soft background
mounts. All filesystems are mounted for read/write access. I suppose the
soft mounts would be more reliable if they were read-only instead. There
are several convenience-versus-reliability tradeoffs with NFS, and you
will have to decide what you are comfortable with.
We strung three local Ethernet trunks down each corridor. This let us do
some load-balancing. We kept the Sun-2s on one of the three trunks, and
we split the Sun-3s across the other two trunks. The three trunks from
each floor went straight to the computer room. There we would take each
pair and connect it to a file server. This way the local Ethernet is no
more loaded than is the file server, typically with ten clients.
Reconfiguring to accommodate each new string of machines is still tough,
though, because usually the new diskless machines get scattered around on
different floors.
[[ Sections on "statistics", "cost", and "hardware" omitted. For the full
text of this message, see the note at the beginning. --wnl ]]
Peter Hill Prime Automation, Inc., Burnsville, Minnesota; (612) 894-0313
UUCP: ihnp4!umn-cs!hall!prime!pfh; sun!tundra!prime!pfh
------------------------------
Date: Thu, 8 Oct 87 11:00:45 EDT
From: peloquin@qtp.ufl.edu
Subject: Re: WYSIWYG editor for the Sun
Regarding the notice on the WYSIWYG editor for sunstations -- the beta
test version of ArborText's Publisher (tm) program is running here on our
system. The system produces beautiful documents on our laser printer, and
provides a handy preview window, which is WYSIWYG with the limitation that
the laser printer has substantially higher resolution than the screen.
The program has facilities for both an equation and a table mode, as well
as for reading in old TeX documents (something we haven't had ocassion to
try out yet). The program has very structured templates for various types
of documents, and will produce beautiful documents -- assuming the
template fits your needs.
The only complaint we have is that the document templates can't be
modified very much too suit our needs, and the whole thing seems to be
a little too mouse-based -- it is annoying, to say the least, to have to
pull out menus 6 deep to make the desired selection. It would also be
nice to allow super- and subscripts in the table mode.
Still, It produces beautiful documents in several fonts and faces, and
many sizes, produces beautiful equations, and allows pictures to be
included from raster files. For those of us who are accustomed to runoff,
troff, wordstar, or Mass-11, this is nothing short of miraculous.
According to ArborText, changes are still being made, so the whole thing
may have improved capabilities and ease of use by the time they are done
with it.
You may want to test the final version out before you decide if it is for
you -- so far we have discovered no really good way to format a paper for
submission to a scientific journal, but are hoping that the next release
with promised improvements will make this possible. Otherwise, I
personally highly recommend it.
Renee Peloquin
peloquin@orange.qtp.ufl.edu
internet adress: [128.227.16.1]
------------------------------
Date: Mon, 12 Oct 87 09:38:49 CDT
From: Michael Begeman <begeman@mcc.com>
Subject: Re: Forking a shelltool from a shelltool
In reply to a question in v5n49 from cosell at BBN...
> That is, the small window pops up, I give the
> host name, type <CR>, the little window goes away... then nothing happens.
> This happens with both 3.2 and 3.4. Is there something obvious or trivial
> I'm doing wrong in the shell script? Is there some workaround that'll do
> what I want? Any help or enlightenment would be appreciated. Thanks
I think you might be running into the same problem that I have. The
rlogin process is a child of the process in the window that pops up to ask
for the host name. When the asking-window goes away, all members of its
process group are killed (in your case, the rlogin process).
I had a sunview program which attempted to let a user send mail from within
it by doing the following:
system("shelltool /bin/csh -c \"mail username\" ");
A window pops up with mail running in it. You compose your message as
usual, type ^D, the mall process exits and the window goes away. But the
mail never gets delivered. Why? Because part of the mail delivery
happens in the "background" after the mail front-end exits, but the
shelltool kills the background part of the mailer when it (the shelltool)
goes away.
[[ Huh? Even if the subprocess has been orphaned? Is it killing the
entre process group? --wnl ]]
I've reported this to Sun. No (useful) response yet. By the way,
this did NOT happen in 3.2. Running mail under "/bin/csh -c" was
enough to save the background process from the ravages of a dying
shelltool back then....
Michael L. Begeman, MCC, 9390 Research Blvd., Austin TX 78759
(512)338-3308 begeman@mcc.com
{gatech,harvard,pyramid,seismo}!ut-sally!im4u!milano!begeman
------------------------------
Date: Wed, 7 Oct 87 21:57:23 CDT
From: wucs1!wucs2!posdamer@uunet.uu.net (Jeff Posdamer)
Subject: DNA product
Yes, we have used the DNA product for six months. It talks to two
microvaxes in our AI lab used to run OPS5. We had little trouble until we
brought up 3.4. We are having messy netmask problems (I think ????) now.
JLPosdamer
------------------------------
Date: Thu, 8 Oct 87 09:45:53 PDT
From: edsel!sears@labrea.stanford.edu (Steve Sears)
Subject: Looking for beta-sites to test Lucid Common Lisp for the Sun 4
Lucid Common Lisp for the Sun 4 will soon be here. We are interested in
locating sites that can provide good feedback by exercising our Sun 4
beta-test version. If you are interested, or know someone that may be,
please reply. You can also reach me by phone at Lucid, (415) 329-8400.
Thanks in advance...
Steve Sears
------------------------------
Date: Thu, 8 Oct 87 15:29:51 EDT
From: lear@aramis.rutgers.edu (eliot lear)
Subject: X PostScript Previewer
I have received a lot of mail asking me where the PostScript Previewer
code can be found. From what was posted to comp.windows.x, there are two
versions available on host berkeley.edu in ~ftp/pub:
ups0.2.vax and ups0.2.sun
There are also .uue versions of the above files in ~uucp.
As our moderator pointed out, most printer ready code would have to be
hacked before being previewed.
Eliot Lear
[lear@rutgers.edu]
------------------------------
Date: Thu, 8 Oct 87 14:42:36 PDT
From: weiser.pa@xerox.com
Subject: Sun-4 bcopy warning
The bcopy routine on the Sun-4's apparently has not been optimized for the
machine at all. Based on the timing, it probably moves only a byte at a
time in a tight loop. An unrolled loop, moving a word at at time (except
for the last few bytes) does much better.
A Sun-4/260 takes 7 seconds to move 10M bytes using repeated calls to
bcopy (8192 bytes/call, all bytes in the cache), but only 1.2 seconds
using a slightly unrolled loop of my own in C. BTW, the latter time is
the same performance as the sun-3/260 gives using bcopy.
-mark
------------------------------
Date: Tue, 6 Oct 87 11:57:27 PDT
From: greg@muddy.cs.unlv.edu (Greg Wohletz)
Subject: Sun csh weirdness
set tmp="aaa"
if($tmp == "bbb") then
if($tmp == "ccc") then
echo ccc
endif
echo bbb
endif
echo aaa
The above csh script echo's ``aaa'' as you would expect when run
under 4.3bsd or ultrix 2.0, however when run under sunos3.4 it echos
``bbb'' and ``aaa'' which it clearly should not.
Somehow the nested if's confuse csh. Any guesses (or facts) as to why?
--Greg
One of the following:
greg@jimi.cs.unlv.edu
<@relay.cs.net:greg@jimi.cs.unlv.edu>
[[ The many mysteries of the c-shell.... Try putting a space between the
keyword "if" and the "(" in the second and third lines (especially the
third line). It worked for me! --wnl ]]
------------------------------
Date: 9 Oct 87 09:19:19 GMT
From: unido!pbinfo!michael@uunet.uu.net (Michael Schmidt)
Subject: varargs: bug or feature?
Investigating <varargs.h>, I ran the program
----------------------------------------------------------------------------
#include <stdio.h>
#include <varargs.h>
varies(va_alist)
va_dcl
{
va_list args;
va_start(args);
if (va_arg(args, int)==0)
printf("0: %d\n", va_arg(args, int));
else
printf("1: %d\n", va_arg(args, short));
va_end(args);
}
main()
{
short sh=1;
int in=1;
varies(0, in);
varies(1, sh);
}
---------------------------------------------------------------------------
and got the following outputs on a Sun (SunOS 3.2),
Pyramid/Targon/35 (OS 3.0) and VAX (4.3 BSD):
Sun:
0: 1
1: 0
Pyramid:
0: 1
1: 0
VAX:
0: 1
1: 1
Well, although the vote is two to one against the VAX, I think the VAX
is write.
Michael Schmidt
UUCP: ...!seismo!unido!pbinfo!michael | Michael Schmidt
or michael@pbinfo.UUCP | Universitaet-GH Paderborn, FB 17
CSNET: michael%pbinfo.uucp@Germany.CSNET | Warburger Str. 100
ARPA: michael%pbinfo.uucp@seismo.css.gov | D-4790 Paderborn, West Germany
------------------------------
Date: 9 Oct 87 22:31:37 GMT
From: franklin@csv.rpi.edu (W. Randolph Franklin ( WRF ))
Subject: Interrupts clog up Sun
Every so often my Sun gets into a stage where it is receiving 3000
interrupts a second, which slows everything else down. Is there any way
to identify where they are coming from and reset it, short of rebooting?
Wm. Randolph Franklin
Preferred net address: Franklin@csv.rpi.edu
Alternate net: wrf@RPITSMTS.BITNET
Papermail: ECSE Dept, Rensselaer Polytechnic Institute,
Troy NY, 12180
Telephone: (518) 276-6077
Telex: 6716050 RPI TROU -- general RPI telex number.
------------------------------
Date: Wed, 07 Oct 87 16:01:27 CDT
From: "Matt Crawford" <matt@oddjob.uchicago.edu>
Subject: SYNC devices on zs or mti ports?
Has anyone connected a synchronous device to a sun through either a zs
port (ttya or ttyb) or an ALM? It seems that either device is
intrinsically capable of operating synchronously. In the case of the zs
ports, the pinouts (from the hardware installation manual) show the
transmit and receive clock signals.
I want to run SLIP through an existing 9600 baud leased line and a pair of
AT&T Dataphone II modems.
Matt Crawford matt@oddjob.uchicago.edu
------------------------------
Date: Wed, 07 Oct 87 17:50:38 MEZ
From: SCHUSTER@DKAUNI13.BITNET
Subject: conversion from IBM/SCRIPT to TeX?
Im not sure if this is the correct list for my question. Maybe you can
help me. My apologies if I'm at the wrong address. One of my colleagues
converted from IBM-world to SunUNIX-world. Now, his problem is the change
of text processing systems. He wants to transfer all his texts from IBM to
UNIX. On IBM he used SCRIPT and GML. Now, he wants to use TeX. My
question is: Does anyone know of a tool which supports the conversion from
SCRIPT/GML to TeX? Every hint is welcome. Thanks in advance Michael
Schuster
-- EMAIL: schuster@dkauni13.bitnet
schuster@germany.csnet
------------------------------
Date: Wed, 7 Oct 87 15:12:45 PDT
From: acad!terry@sun.com (Terry Fritz ext.636)
Subject: Printcap entry for HP-Laserjet?
Can somebody send me a printcap for a HP-laserjet for a Sun?
Thanks,
Terry Fritz
------------------------------
Date: 8 Oct 87 22:14:13 GMT
From: ll-xn!atexrd!sda@rutgers.edu (Stephen Ayers)
Subject: X.25 using Suns?
At Atex, Inc. we are looking into connecting our suns to X.25 for overseas
communications. It will be used mostly for uucp/mail. Any information on
setting up such a link would be greatly helpfull. Including L-devices,
L.sys and modifications to the uucp source.
Thanks in advance!
Steve Ayers, Atex, Inc., A Kodak Company
{ll-xn,genrad,kodak,munsell}!atexrd!sda
+1 617 276-7384
------------------------------
Date: Tue, 06 Oct 87 14:27:01 -0700
From: Bob Brown <rlb@riacs.edu>
Subject: Mandelbrot icon
For all of us who like to draw Mandelbrot sets, here's an icon to start
with, generated computationally.
Bob
/* Format_version=1, Width=64, Height=64, Depth=1, Valid_bits_per_item=16
*/
0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
0x0000,0x0000,0x0000,0x8000,0x0000,0x0000,0x0001,0xC000,
0x0000,0x0000,0x0003,0xE000,0x0000,0x0000,0x0003,0xF000,
0x0000,0x0000,0x0003,0xE000,0x0000,0x0000,0x0001,0xC000,
0x0000,0x0000,0x002F,0xFA00,0x0000,0x0000,0x033F,0xFE00,
0x0000,0x0000,0x03FF,0xFFB0,0x0000,0x0000,0x01FF,0xFFF0,
0x0000,0x0000,0x07FF,0xFFE0,0x0000,0x0000,0x1FFF,0xFFF0,
0x0000,0x0000,0x1FFF,0xFFF0,0x0000,0x0000,0x3FFF,0xFFFC,
0x0000,0x0000,0x3FFF,0xFFFC,0x0000,0x0020,0x3FFF,0xFFF8,
0x0000,0x02F0,0x7FFF,0xFFF8,0x0000,0x03FC,0x7FFF,0xFFF8,
0x0000,0x0FFE,0x7FFF,0xFFF8,0x0000,0x0FFF,0x7FFF,0xFFF8,
0x0000,0x0FFF,0x7FFF,0xFFF0,0x0000,0x6FFF,0x7FFF,0xFFF0,
0x7FFF,0xFFFF,0xFFFF,0xFFC0,0x0000,0x6FFF,0x7FFF,0xFFF0,
0x0000,0x0FFF,0x7FFF,0xFFF0,0x0000,0x0FFF,0x7FFF,0xFFF8,
0x0000,0x0FFE,0x7FFF,0xFFF8,0x0000,0x03FC,0x7FFF,0xFFF8,
0x0000,0x02F0,0x7FFF,0xFFF8,0x0000,0x0020,0x3FFF,0xFFF8,
0x0000,0x0000,0x3FFF,0xFFFC,0x0000,0x0000,0x3FFF,0xFFFC,
0x0000,0x0000,0x1FFF,0xFFF0,0x0000,0x0000,0x1FFF,0xFFF0,
0x0000,0x0000,0x07FF,0xFFE0,0x0000,0x0000,0x01FF,0xFFF0,
0x0000,0x0000,0x03FF,0xFFB0,0x0000,0x0000,0x033F,0xFE00,
0x0000,0x0000,0x002F,0xFA00,0x0000,0x0000,0x0001,0xC000,
0x0000,0x0000,0x0003,0xE000,0x0000,0x0000,0x0003,0xF000,
0x0000,0x0000,0x0003,0xE000,0x0000,0x0000,0x0001,0xC000,
0x0000,0x0000,0x0000,0x8000,0x0000,0x0000,0x0000,0x0000,
0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000
[[ Stored in the archives as "sun-icons/rlb-mandelbrot.icon". --wnl ]]
------------------------------
End of SUN-Spots Digest
***********************