[comp.sys.sun] avoiding obsolescence

mc@moc.jpl.nasa.gov (Mike Caplinger) (11/15/89)

Reading this list, one frequently sees exhortations to buy more memory,
get a faster processor, or otherwise totally replace one's hardware in
order to "keep up" with the latest SunOS release.  A recent message, for
example, speculated that 12 MIPS and 16 Meg was going to be the minimal
usable configuration for SunOS/Xview, and my experience so far supports
that guess.  Some of us, however, can't afford the luxury of trading up to
a new CPU every 18 months.

Sure, one can stick at some release of SunOS, like all the 3/50 users more
or less sticking at 3.5, but neither Sun nor third-party software vendors
makes this very desirable.  Would it be remotely possible for Sun to
continue maintaining release M.X when M+1.X was out in the marketplace?
(Given current trends, maybe that should be release 4.M and 4.M+1, or even
release 4.0.M and 4.0.M+1 ;-)  Wouldn't it be nice if only they could?
Are there enough forces in the marketplace to make this kind of parallel
development possible?  What if some third party did it?

The sad thing is, I remember when a 2-meg 68010 Sun configuration was
usable, and even though I have a 20-meg Sun 4 now, I don't think I've
gotten an order of magnitude better performance, from a whole-system
perspective.

	Mike Caplinger, ASU/Caltech Mars Observer Camera Project

p.s.  Maybe all the software developers at Sun should be given Sun 2s, and
made to use them.

davecb@nexus.yorku.ca (David Collier-Brown) (11/21/89)

miranda!mc@moc.jpl.nasa.gov (Mike Caplinger) writes:
| Reading this list, one frequently sees exhortations to buy more memory,
| get a faster processor, or otherwise totally replace one's hardware in
| order to "keep up" with the latest SunOS release. [...]
| Some of us, however, can't afford the luxury of trading up to
| a new CPU every 18 months.

| The sad thing is, I remember when a 2-meg 68010 Sun configuration was
| usable, and even though I have a 20-meg Sun 4 now, I don't think I've
| gotten an order of magnitude better performance, from a whole-system
| perspective.

Well, this raises a relate problem I'm looking at: the widening of the
"market" for workstations in our organization.

We have administrative users who could make use of a windowing
environment, primarily for keeping track of **many many** simultaneous
concerns/tasks.  We have support (secretarial) users who would love to be
able to do more than a single task at a time.

Both of these types of people can be "sold" machines like 3/50s. But a
3/50 (and my old 2/50) cannot be provided with enough memory to usefully
run SunOS 4...  Which means we freeze them at 3.5, and hope they wear out
before Sun drops support.

So here's a question for Sun and for the readers of this list: how do we
broaden the customer base?  How long **should** we keep a machine around
after the power-freak developers have got a faster one?  How long can we
get hardware and software support for one?

David Collier-Brown,  | davecb@yunexus, ...!yunexus!davecb or
72 Abitibi Ave.,      | {toronto area...}lethe!dave 
Willowdale, Ontario,  | Joyce C-B:
CANADA. 416-223-8968  |    He's so smart he's dumb.

yar@basser.cs.su.oz.au (11/24/89)

In article <3103@brazos.Rice.edu> miranda!mc@moc.jpl.nasa.gov (Mike Caplinger) writes:
> X-Sun-Spots-Digest: Volume 8, Issue 199, message 14 of 17
>
> Reading this list, one frequently sees exhortations to buy more memory,
> get a faster processor, or otherwise totally replace one's hardware in
> order to "keep up" with the latest SunOS release.  A recent message, for
> example, speculated that 12 MIPS and 16 Meg was going to be the minimal
> usable configuration for SunOS/Xview, and my experience so far supports
> that guess.  Some of us, however, can't afford the luxury of trading up to
> a new CPU every 18 months.

Is anyone out there working on a native version of an Xserver for sun
3/50s so that they can be used as very nice X-terminals, no SUNOS, no
paging, no need for more memory. That way we can find a cheap way of
getting out of this expensive workstation architecture and back to a
system where you replace the expensive back-end when it runs out of grunt
(every 2-3 years) load it full of memory and disks and keep the user
interfaces (19" windowing terminals) for about 10 years, like we were able
to with RS232 ttys. We have been wanting to do this for our own 3/50s for
years but are still waiting for our sunos source, but then again it has
only been 4 years so who is rushing????  I'd rather have 30 19" Xterminals
backing onto a couple of Mips R6000 based systems each with 256Mb of
memory than have 30 Sparcstations with each with the minimum 16Mb of
memory.  It would sure make the management a lot easier, and it would
eliminate the factor of 30 involved in any upgrade.

Just as I was actively thinking of dumping Sunos3.5 and buy some memory
someone from sun was advising a minimum of 16Mb or you will be stuck with
Sunos4.0.?, where will it end????

Ray Loyzaga
Basser Department of Computer Science
Sydney University
yar@cs.su.oz

wcs@ho95c.att.com (William Clare Stewart) (12/10/89)

Dave Collier-Brown wrote about the problem of Sun 3/50's being obsolete -
not enough memory to run 4.0.X, but not wanting to stick the users with
3.5 either.  The throw-money-at-it solution is to buy upgrade boards for
about $1000 ($500=board + 4MB SIMM), which will give you 8MB - enough to
run 4.0.X or 3.5 with lots of windows.  (The person who answered the phone
at one of the memory vendors said she's totally non-technical and had no
trouble installing it herself.)

In article <3319@brazos.Rice.edu> yar@basser.cs.su.oz.au writes:
> Is anyone out there working on a native version of an Xserver for sun
> 3/50s so that they can be used as very nice X-terminals, no SUNOS, no
> paging, no need for more memory. That way we can find a cheap way of

When I last talked with Grasshopper Group this summer, they were working
on a NeWS server for the 3/50, which does the same kind of thing, in
good-looking Postscript (instead of clunky rasters.  The demo I saw at
Usenix looked good, and they expected to have a version that never needed
to page by the time they finished.  This will let you use your 3/50s as
the execllent cheap 19" system they are, while buying some real computing
power for the back-end.

mjacob@sun.com (Matt Jacob) (12/21/89)

There is no modern, Unix based, 32 bit virtual memory paging system that
works in under 8mb. By modern, I mean all the covergence of both BSD and
System V crap that everyone, even University folks, want. NeXT couldn't do
it. Neither could DEC.

I am not pleased with our performance viz. memory, but I can't do too much
about it. 4.1 (you didn't hear it from me) will be a tad better than
4.0.X, but the *glory* days of 3.5 may not be acheived.

I'd hate to lose you as a customer, but if you like, I'll sell you my home
512kb LSI-11/23 system (runs 2.9 BSD). It is also slower 'n snot.  

-matt jacob