Very Computer

by **Mark McWiggi** » Wed, 30 Mar 1994 18:48:22

PC Week this week says that part of the agreement for Sun to join OSF
was for OSF to stop development on OSF/1, DCE, and DME. This is in the
Spencer Katt rumors column, and I don't know if he's reliable or
Dvorakish ...

--
Mark McWiggins Hermes & Associates +1 206 632 1905 (24 hrs.)

by **Joseph_Jack..** » Thu, 31 Mar 1994 00:00:06

> PC Week this week says that part of the agreement for Sun to join OSF
> was for OSF to stop development on OSF/1, DCE, and DME. This is in the
> Spencer Katt rumors column, and I don't know if he's reliable or
> Dvorakish ...

I started to wonder about PC Week back when they explained that remote
procedure call technologies are layered on top of distributed file
system technologies. I've always been taught that it's the opposite!
;-) I wish I still had that article to quote, but I tossed it. I'll
have to go check out this week's issue.

Joe Jackson,
File Systems Product Support,
Transarc Corp.

by **Douglas Jon** » Fri, 01 Apr 1994 02:43:38

........deleted............

Quote:
>OSF/1 would seem to be a less serious competitor than Solaris 2.x,
>at this moment. Solaris 2.x is not supported by Sun alone. It
>is supported by a number of other vendors as well, including Amdahl
>and CRAY. Solaris is also available for a larger number of platforms,
>at this moment. Though OSF/1 may be a good platform, it seems most
>likely that a future unified Unix is based on System Vr4 + COSE,
>which is pretty much what Solaris will look like in future.
>Some of OSF's products will be integrated in future products
>of COSE. But it is unlikely that OSF/1 will survive much longer,
>especially condering that DEC is now ranking fifth with less than 10%
>of the Unix market share (in $$).

>Casper

Isn't there a new Cray out (t3d or something) that useing the ALPHA processors?
If so, will this have any impact on OSF/1?

Cheers,
Douglas

by **Jim Hog** » Fri, 01 Apr 1994 05:33:24

>........deleted............

>> [...trimmmmm ...]
>>of COSE. But it is unlikely that OSF/1 will survive much longer,
>>especially condering that DEC is now ranking fifth with less than 10%
>>of the Unix market share (in $$).

>>Casper

As someone in\volved in an evaluation of OSF/1 (among other Unix
possibilities), this seems like a rather sweeping statement.
On what specifics do you base "unlikely that OSF/1 will survive
much longer"? No begging an argument, but am truly interested.
LAck of support or negative indications from DEC? Lack of
adopters?

Quote:>Isn't there a new Cray out (t3d or something) that useing the ALPHA processors?
>If so, will this have any impact on OSF/1?
>Cheers,
>Douglas

Jim

by **Casper H.S. D** » Fri, 01 Apr 1994 16:59:07

>Isn't there a new Cray out (t3d or something) that useing the ALPHA processors?
>If so, will this have any impact on OSF/1?

Current OSF/1 offerings do not run on multi-processor machines.
OSF/1 2.0 probably will, but isn't there yet.

Experience of otehr vendors says that it is that supporting
MP isn't enough to make it scale well. Sun now ships 20-WAY
SC2000s, which seem to work for some applications.

OSF/1 2.x will support MP, I think, but they will probably have the
problems as Sun and otehr had. First you get it to work on 2 way machine
and then you have a long way to go to get most out of N-way systems.

Casper

by **Jeff Cart** » Sat, 02 Apr 1994 01:26:14

Quote:> Current OSF/1 offerings do not run on multi-processor machines.
> OSF/1 2.0 probably will, but isn't there yet.

> Experience of otehr vendors says that it is that supporting
> MP isn't enough to make it scale well. Sun now ships 20-WAY
> SC2000s, which seem to work for some applications.

That should read "current _DEC_ OSF/1 offerings do not run on multiprocessor
machines". OSF/1 R1.0 and R1.1 (from OSF) ran on the Encore Multimax (routinely
used 8-way in-house, tested 12-way). This was productized by IBM/Kingston as
AIX/ESA, (an MP system) and Kendall Square Research as KSR/1. The OSF/1 R1.2
MP reference platform was the was the Sequent Symmetry. We typically use 4 to 8
CPU machines, but they go up to about 20. The MP scaling in this size range
is well understood, and the base technology has been shown to be 'scalable'.
I can't comment on DEC's plans & schedules, but there is no reason not to
expect it to do 'as well' as other MP systems.

jeff carter

by **Glenn Holling** » Sat, 02 Apr 1994 01:29:05

: >Isn't there a new Cray out (t3d or something) that useing the ALPHA processors?
: >If so, will this have any impact on OSF/1?

: Current OSF/1 offerings do not run on multi-processor machines.
: OSF/1 2.0 probably will, but isn't there yet.

: Experience of otehr vendors says that it is that supporting
: MP isn't enough to make it scale well. Sun now ships 20-WAY
: SC2000s, which seem to work for some applications.

: OSF/1 2.x will support MP, I think, but they will probably have the
: problems as Sun and otehr had. First you get it to work on 2 way machine
: and then you have a long way to go to get most out of N-way systems.

I think you are being overly pessimistic about DEC's abilities in the
MP feild. DEC has lots of experience with up to 6-way (and more in
the lab) multiprocessing on VMS. The operating system is different,
but many of the issues are the same. Note that DEC is delivering
Symetric MP with OSF/1 2.0, which scales much better than the
Master/Slave version supported in some Ultrix versions on the VAX
11/782.

other things are real issues, including lack of adoption by other
vendors and (apparent) waning interest in OSF to take OSF/1 forward.

Glenn Hollinger,
University of Saskatchewan.

by **unix** » Sat, 02 Apr 1994 01:31:53

>Isn't there a new Cray out (t3d or something) that useing the ALPHA processors?
>If so, will this have any impact on OSF/1?

I don't know how it will impact OSF/1, but you were correct. The Cray
T3D does use Digital's Alpha, but it is an MP add-on to your YMP.
Currently, they have a 1024 processor model out, and are planning on
2048 to 8192 sometime soon. Wow! That will blow away the other
MP machines. I prefer the hyper-torodial design to the mesh or
hypercube -- no more furthest-neighbor problems.

--

by **Eric D Osbor** » Sat, 02 Apr 1994 16:51:58

I want to tread carefully here, so I don't sound like a commercial, but I
think much ado is being made about nothing relative to the the DEC-is-the-
only-OSF/1-user thing. First, Digital has a huge engineering staff assigned
to OSF/1, and is NOT dependent upon OSF for future maintenence of the product.
Secondly, Digital is not different in any way from the other vendors in this
issue - Solaris has a SVR4 base, but has many SUN enhancements - further,
Sun has recently made it clear (via a $82 million buyout of licensing rights)
that they want absolute freedom to do as they will with Solaris. In the
final analysis, Digital has DEC/OSF1, Sun has Solaris, HP has HP/UX, etc. -
and each of those companies is the sole support arm for those products.

For all of this, there are phenomemally convoluted cross-licensing agreements
between the various companies. Virtually all the major companies license
critical technology to the others. If the right things are done, the
products conply with standards, and have good compatibility.

As Bob Palmer recently said at Uniforum,
it is standards compliance that matters, and if customers believe that then
customers will spend their dollars with vendors that provide products with
those standards. Digital does as good a job there as any other vendor, and
better than most. Digital is a participant in the COSE processes, and is
a significant technology contributor. Digital is also actively involved in
standars efforts with XOPEN, ANSI, and a myriad of other places.

So, DEC OSF/1 isn't going away. Digital is investing massive amounts of money
and resources in it - certainly on a scale similar to the investments in
Solaris or AIX or HP/UX, and probably larger in some cases. If you have
problems with individual applications vendors, it's not always Digital's
fault - sometimes those vendors have agendas of their own. One way
to help insure that your applications of choice are available where you
want them is to make it clear to the applications vendor that not supporting
your chosen platforms will make you look for applications that do support the
hardware you want. Digital has a very aggressive program in effect to get
ports on OSF/1 - but we can't always convince the software companies. As long
as you, as customers, are willing to allow an application vendor to coerce you
into buying hardware you don't want, they have no incentive to port to our, or
any other platform. It's a matter of simple economics - if they're going to
get the sale no matter what, why incur the cost of a port?

Vote with your dollars. Reward vendors who give you what you want, both
hardware and software. The open systems market has made it clear that
customers don't wan't to be held hostage by hardware vendors with
non-competitive pricing. If the marketplace allows software vendors to
dictate hardware choices, that choice in hardware becomes irrelevant, and
consumers will ultimately see a reduction in competition and an increase
in prices.

Sorry about the length.

Eric Osborne

===========================================================================
The opinions expressed above are those of the author, and do not necessarily
represent those of Digital Equipment Corporation.

===========================================================================

by **Casper H.S. D** » Sat, 02 Apr 1994 19:04:51

>I think you are being overly pessimistic about DEC's abilities in the
>MP feild. DEC has lots of experience with up to 6-way (and more in
>the lab) multiprocessing on VMS. The operating system is different,
>but many of the issues are the same. Note that DEC is delivering
>Symetric MP with OSF/1 2.0, which scales much better than the
>Master/Slave version supported in some Ultrix versions on the VAX
>11/782.

It seems that OSF/1 does in fact run on MP systems. The MP reference
port is a Sequent Symmetry, I'm told. Apparently DEC did have other
priorities than making it run on MP alphas. One reason might be that
they'd rather wait for OSF/1 2.0, so they'd have to port it only once.

Quote:>other things are real issues, including lack of adoption by other
>vendors and (apparent) waning interest in OSF to take OSF/1 forward.

The latest development seems to be that OSF will continue to exist as
organisation and work in sort-of the same fashion, but with less
in-house development, but that it will be renamed as soon as someone
finds a better name. The rumour mill says that part of this
restructuring is the end of some of the OSF/1 technologies,
OSF/1 and DCE being named as two of them. But that's just a rumour.

Casper

by **Jay Mayna** » Sat, 02 Apr 1994 20:44:26

Quote:>it is standards compliance that matters, and if customers believe that then
>customers will spend their dollars with vendors that provide products with
>those standards.

Indeed. We just bought an Alpha 3000/400 because OSF/1 was much closer to
standard in the original, unaltered setup on the machine than Solaris 2.3.

Quote:>If you have
>problems with individual applications vendors, it's not always Digital's
>fault - sometimes those vendors have agendas of their own.

Well, maybe. We've had at least one vendor tell us that they wanted to port
to OSF/1, but that DEC wouldn't help them unless they ported to OpenVMS
first. They, and I, think that's astoundingly short-sighted. (So did my
local DEC rep when I related the story, leading me to believe that the left
hand probably doesn't know what the right hand is doing.) We don't think that
system is critical to our shop, so it didn't get in the way of our
purchase...but if it's true, it's disturbing.

(Disclaimer: I don't work for the University of Denver; I'm in Houston.)

Quote:> One way
>to help insure that your applications of choice are available where you
>want them is to make it clear to the applications vendor that not supporting
>your chosen platforms will make you look for applications that do support the
>hardware you want.

This, however, is exactly backwards from just about everyone's recommendation
on selecting a computer system...the normal recommendation is to select the
application you want, and then select the system to run it on. That usually
produces much better results overall where it counts - at the user level. Of
course, if you're doing open systems, the platform can change while the same
software sticks around. That we're not there yet does nothing more than point
out the fallacy of open systems.

Quote:> Digital has a very aggressive program in effect to get
>ports on OSF/1 - but we can't always convince the software companies.

...and, if the story above is true, I can see why.

(Wishing I had a keyboard with a real escape key on my VT220 in VT220 mode...)
--
Jay Maynard, EMT-P, K5ZC, PP-ASEL | Never ascribe to malice that which can

"The difference between baseball and politics is that, in baseball, if you
get caught stealing, you're out!" -- Ed Shanks

by **Jim Bals** » Sun, 03 Apr 1994 00:06:46

Sun is a very influential company. But not that influential. The work on
OSF/1 will stop because more than anything else, DEC is the only company using
the beast. Why should HP, IBM, SUN, NOVELL, and anyone else's yearly payments
into the OSF/ORG94 pay for DEC's OS technology.

Work on DME will most end also, but not because SUN said so. DME has
lost momentum because it is so late to market.

My opinions only.

Jim

by **Donald Linds** » Tue, 05 Apr 1994 05:08:51

Quote:>OSF/1 2.x will support MP, I think, but they will probably have the
>problems as Sun and otehr had. First you get it to work on 2 way machine
>and then you have a long way to go to get most out of N-way systems.

Sun added MP support "recently".

OSF/1 started with the Mach kernel, developed here at CMU. At the
time OSF was founded, we already had Mach running N-way, N>2.

--
Don D.C.Lindsay Carnegie Mellon Computer Science

Very Computer

OSF/1 vs. Solaris