[comp.sys.dec] Academic workstations

cullum@lclark.UUCP (Mike Cullum) (06/09/89)

Greetings:

We are in the process of considering the purchase of workstations for
a small lab in our Computer Science Department.  Our proposed 
configuration calls for 8 workstations (8Mb RAM, 200+Mb disk, large
monochrome display) and a server.  

We are vacillating between Apple AUX, NeXt, and Suns.   What are
people using?  Is there anyone who is using Apple AUX in a
lab situation regularly?  Anyone using the DecStation 3100?
Any advice?

Thanks in advance for the help.

Mike Cullum
Lewis & Clark College
Portland, Oregon

UUCP: tektronix!reed!lclark!cullum
Bitnet: cullum@lclark

pauls@apple.com (Paul Sweazey) (06/09/89)

In article <507@lclark.UUCP> cullum@lclark.UUCP (Mike Cullum) writes:
> We are vacillating between Apple AUX, NeXt, and Suns.   What are
> people using?  (Etc.)
> 
> Thanks in advance for the help.
> 
> Mike Cullum

Everyone here says I should tell you to use Apple AU/X.  :-)
pauls

Paul Sweazey
Apple Computer, Inc.
pauls@apple.com
(408)-974-0253

jtwarden@pawl.rpi.edu (Joseph T. Warden) (06/10/89)

In article <2302@internal.Apple.COM> pauls@apple.com (Paul Sweazey) writes:
>In article <507@lclark.UUCP> cullum@lclark.UUCP (Mike Cullum) writes:
>> We are vacillating between Apple AUX, NeXt, and Suns.   What are
>> people using?  (Etc.)
>> 
>> Thanks in advance for the help.
>> 
>> Mike Cullum
>
>Everyone here says I should tell you to use Apple AU/X.  :-)
>pauls
>
>Paul Sweazey
>Apple Computer, Inc.

Another opinion (my own) is to go with the Suns - you have access
to a large volume of software (PD, etc), a large installed base
(esp. in Academia) and good pricing. An alternative is DEC, but
if you want to work with server/clients, I think Sun is probably
the easiest to implement and maintain. This opinion is from a
chemist, whose philosophy is to extract the greatest use from
the computer without being consumed by the process.

Joseph Warden
Renssealer
jtwarden@pawl.rpi.edu

abstine@sun.soe.clarkson.edu (Arthur Stine) (06/10/89)

Although workstation speed has increased dramatically over the past few years
(Sun, DEC, Apollo, etc all now have machines which run in excess of 5 mips and
the newer RISC technologies push this up to and past 15mips), the speed of the
local area networks have not (Ethernet is still widely used, with token rings
seeing some increasing use). One cannot realize the maximum potential of the
higher performance workstation by using them as a diskless workstation.

Here at Clarkson University, we have a number of Sun workstations (3/50's, 
3/60's, 386i's) which are served off of a couple of 3/260 servers. The 3/50's
are primarily diskless, but the others all have local disk. We also have
15 Vaxstations (all diskless except for 1 with local page/swap) served off
of a vaxserver-3500. There are additional Suns in other departments which
are the same sort of technology (3/50's served from a 3/260). Performance
is adaquete for the diskless workstations, but when the network becomes
heavily used, the diskless stations feel the pinch. 

Overall, the best bet is probably to use central servers for user files,
large applications, etc and equip the stations with local page/swap and
their own set of systems files (this applies to both Unix and VMS systems,
as the both will make heavy use of the network for I/O in the client/server
setup). This way, you can have the users files residing in a common place, 
accessible from a number of places, but still realize the high performance of
the workstations. Making a DECstation or Sparcstation page/swap and do all
of its I/O across an Ethernet will not make the workstation seem very fast.
And it won't take many of the faster workstations to load down the net. Note
also that the RISC systems generally have larger images than their CISC
counterparts.

Diskless only stations (in my opinion) are the wave of the past. Without
higher bandwidth networks, they have become throttled by the available
network technology.

art stine
sr network engineer
clarkson u

burzio@mmlai.UUCP (Tony Burzio) (06/12/89)

In article <5386@rpi.edu>, jtwarden@pawl.rpi.edu (Joseph T. Warden) writes:
> In article <2302@internal.Apple.COM> pauls@apple.com (Paul Sweazey) writes:
> >In article <507@lclark.UUCP> cullum@lclark.UUCP (Mike Cullum) writes:
> >> We are vacillating between Apple AUX, NeXt, and Suns.   What are
> >> people using?  (Etc.)
> >Everyone here says I should tell you to use Apple AU/X.  :-)
> 
> Another opinion (my own) is to go with the Suns - you have access
> to a large volume of software (PD, etc), a large installed base

I personally like Hewlett Packard, the number one workstation vendor.  Other
people here like Suns and Apples (they make pretty pictures).  All three
fit in pretty well together.  The only vendor I would stay away from is that
other three letter initial vendor who sells (ugh) proprietary operating
systems... I think it is CED or EDC or something... :-)

*********************************************************************
Tony Burzio               * Do humanity a favor,
Martin Marietta Labs      * shoot a Communist...
mmlai!burzio@uunet.uu.net *
*********************************************************************

barnett@crdgw1.crd.ge.com (Bruce G. Barnett) (06/13/89)

In article <3160@sun.soe.clarkson.edu>, abstine@sun (Arthur Stine) writes:
>
>One cannot realize the maximum potential of the
>higher performance workstation by using them as a diskless workstation.

I question this statement. See below.

>Here at Clarkson University, we have a number of Sun workstations (3/50's,
>3/60's, 386i's) which are served off of a couple of 3/260 servers. The 3/50's
>are primarily diskless, but the others all have local disk.

I would say you have your machines configured backwards.
The 3/50's need the disk because they are the ones most likely
to page, and if they have local disks, they won't flood the network.

>Performance
>is adaquete for the diskless workstations, but when the network becomes
>heavily used, the diskless stations feel the pinch.

I would say you have a network configuration problem.

>Diskless only stations (in my opinion) are the wave of the past. Without
>higher bandwidth networks, they have become throttled by the available
>network technology.

I will agree that an application that has to work with a lot of data
(Image processing, etc) will need a fast disk.

We have 300 Sun's here, and 80-90% are diskless.  They are all on a single
Ethernet (with several seqments and smart bridges).

If a machine is slow,  we add more memory. If we can't (like a Sun 3/50),
we try to add a local SCSI disk to eliminate the swaping.

Some high end servers have their own disk, that's true.
But with the coming of network window systems, and diskless machines with 64
megabytes of memory, I do not agree with your statement about diskless
machines being passe.

In our experience, we see a continued use of diskless workstations.
If we had to convert over to diskfull workstations, we would have to
double or triple the support staff needed to keep them all running.

--
Bruce G. Barnett	<barnett@crdgw1.ge.com>  a.k.a. <barnett@[192.35.44.4]>
			uunet!crdgw1.ge.com!barnett barnett@crdgw1.UUCP

raveling@venera.isi.edu (Paul Raveling) (06/24/89)

In article <5386@rpi.edu> jtwarden@pawl.rpi.edu (Joseph T. Warden) writes:
>
>Another opinion (my own) is to go with the Suns - you have access
>to a large volume of software (PD, etc), a large installed base
>(esp. in Academia) and good pricing. An alternative is DEC, but
>if you want to work with server/clients, I think Sun is probably
>the easiest to implement and maintain. This opinion is from a
>chemist, whose philosophy is to extract the greatest use from
>the computer without being consumed by the process.

	Since I'm now involved with supporting workstations here,
	I'll offer a relative rating of software/system quality
	as I see it:

		Best:	HP
		#2:	DEC
		Worst:	Sun

	HP & DEC are probably close, but we have mainly Sun & HP
	workstations; I don't have much DEC experience to confirm
	this suspicion.  Sun's software (e.g., C compiler) is often
	visibly less refined and more trouble-prone than HP's.

	BTW, these are my opinions based on my experience.
	Milage may vary for others...


----------------
Paul Raveling
Raveling@isi.edu

fair@apple.com (Erik E. Fair) (06/25/89)

In article <8732@venera.isi.edu> raveling@venera.isi.edu (Paul Raveling) 
writes:
>     HP & DEC are probably close, but we have mainly Sun & HP
>         workstations; I don't have much DEC experience to confirm
>         this suspicion.  Sun's software (e.g., C compiler) is often
>         visibly less refined and more trouble-prone than HP's.

Just try getting Ultrix source from DEC. When you do, you'll find out that 
it is "for reference purposes only."

        Erik E. Fair      apple!fair      fair@apple.com

raveling@venera.isi.edu (Paul Raveling) (06/27/89)

In article <2467@internal.Apple.COM> fair@apple.com (Erik E. Fair) writes:
>In article <8732@venera.isi.edu> raveling@venera.isi.edu (Paul Raveling) 
>writes:
>>     HP & DEC are probably close, but we have mainly Sun & HP
>>         workstations; I don't have much DEC experience to confirm
>>         this suspicion.  Sun's software (e.g., C compiler) is often
>>         visibly less refined and more trouble-prone than HP's.
>
>Just try getting Ultrix source from DEC. When you do, you'll find out that 
>it is "for reference purposes only."

	We haven't managed to get HP-UX source from HP either.
	Having viewed things like this from the vendor side as well
	as from the user side, I can't blame them much for not wanting
	wanting to maintain a source product.

	Instead I just mutter sometimes that a good operating system 
	shouldn't leave users with a need for its source.  Unfortunately
	that kind of need seems to keep arising on all the variants of UNIX
	I've dealt with.


----------------
Paul Raveling
Raveling@isi.edu

rec@dg.dg.com (Robert Cousins) (06/27/89)

In article <8767@venera.isi.edu> raveling@venera.isi.edu.UUCP (Paul Raveling) writes:
>In article <2467@internal.Apple.COM> fair@apple.com (Erik E. Fair) writes:
>>Just try getting Ultrix source from DEC. When you do, you'll find out that 
>>it is "for reference purposes only."
>	We haven't managed to get HP-UX source from HP either.
>Paul Raveling
>Raveling@isi.edu

Having watched this thread for a while now, I felt that I just COULDN'T
watch any more.  First of all, when you want an academic workstation, 
you sould go for a standard which allows you to buy compatible hardware
from multiple vendors and still run the same binaries. Secondly, any
vendor which will not sell the source code for their operating system
must have something to hide.  The DG/UX sources are licensable and have
been licensed by several third parties and currently is being offered
for resale on their hardware.  

When I was in school, the real issue with budgets was to get the most
bang on very few bucks.  I recommend that you check into some of the 
new 88K products (from a number of companies aside from my own).  These
are faster than the machines discussed in this thread earlier AND cost
much less.  In some cases under $500/MIPS ready to run.

To avoid turning this into a commercial for my product, I will close
here.  If anyone is interested in 88K based products, just let me know.

Robert Cousins
Dept. Mgr, Workstation Dev't.
Data General Corp.

Speaking for myself alone.
 

avolio@decuac.dec.com (Frederick M. Avolio) (06/28/89)

ULTRIX sources are most certainly available from Digital and they
are the same sources that are used to build the binaries.  A handful of
code is not generally available -- VAX/C, LAT code, etc -- but most of
it is there and orderable.

Fred

bukys@cs.rochester.edu (Liudvikas Bukys) (06/28/89)

In article <197@dg.dg.com> uunet!dg!rec (Robert Cousins) writes:
>
>Having watched this thread for a while now, I felt that I just COULDN'T
>watch any more.  First of all, when you want an academic workstation, 
>you sould go for a standard which allows you to buy compatible hardware
>from multiple vendors and still run the same binaries. Secondly, any
>vendor which will not sell the source code for their operating system
>must have something to hide.  The DG/UX sources are licensable and have
>been licensed by several third parties and currently is being offered
>for resale on their hardware.  
>
>When I was in school, the real issue with budgets was to get the most
>bang on very few bucks.  I recommend that you check into some of the 
>new 88K products (from a number of companies aside from my own).  These
>are faster than the machines discussed in this thread earlier AND cost
>much less.  In some cases under $500/MIPS ready to run.
>
>To avoid turning this into a commercial for my product, I will close
>here.  If anyone is interested in 88K based products, just let me know.
>
>Robert Cousins
>Dept. Mgr, Workstation Dev't.
>Data General Corp.

Speaking as someone who has just been through a long complicated
workstation purchase decision, I have the following things to say:

We took competitive bids from all the major workstation vendors, and I
must say that there were only two companies (and neither of them is DG)
with serious bids from the price/performance point of view.  BTW,
competition is wonderful!

A number of people here had high hopes for the DG product, but all
these fast and loose $/MIPS figures that look so good are for seriously
underconfigured machines.  Add in appropriate amounts of memory and
maybe a local paging disk, and things don't look so good any more.
(The design/win beta prices looked better, but those aren't the current
prices.)

DG needs to realize that it is a relative unknown in this field, and
will need to prove itself in the next year or two.  Let me tell you how
this all sounds to a customer (who gave DG's product serious thought
but clearly made the right decision buying something else):

DG says: "DG/UX is great, new reliable filesystem, multi-processor support."
Customer thinks: "DG has no reputation as a Unix vendor yet.  And just what we
	needed, another vaguely incompatible Unix clone with underlying system
	stuff that may or may not work well, and probably with a whole new
	incompatible set of sys admin tools to support it."

DG says: "$/MIP."
Customer thinks:  "We're trying to get rid of our 4-MB machines and this vendor
	thinks we should buy more.  And their memory is on these 4MB cards that
	nobody else makes yet, so I pay through the nose for more, if I decide
	I want a real machine."

DG says: "ABI"
Customer thinks: "AT&T and SUN thought of it first and I can buy Suns and Solbournes
	(both machines that I might want) now.  And there's a whole heck of a lot
	more software for SUNs that I can buy right now.  Also, there is a sizable
	user community for SUNs.  Who the heck owns DGs yet?"

DG says "88K"
Customer thinks: "I don't know who's going to win in the end; the vendors will be
	leapfrogging each other for quite a while and whatever I buy won't be
	`the best' for very long so I'd better not lose any sleep over it."

DG says "Available now."
DG salesman says "Can you delay your purchasing decision for a month?"
Customer says "Product not ripe.  Also, Motorola tends to pre-announce A LOT,
	and they haven't announced anything (other than DG's ECL implementation)"

My point is that a relative unknown like DG is breaking into a market where they
have not had a product before, and in which they have no reputation.  From my
point of view, they need to do as little pre-announcing as possible, and just
deliver the product so people can try it out.  Also, they should minimize the
touting of benefits that are really only theoretical right now, like the 88K ABI.

annala@neuro.usc.edu (A J Annala) (07/01/89)

>I have been doing traffic analysis, and it is our diskless
>workstations which are pulling down our network.  Our network
>throughput is starting to drop because of the high collision rate,
>BUT...
>
Maybe you should consider partitioning your network into a group of smaller
branches using intelligent ethernet bridges to reduce traffic flow from one 
part of the network onto other parts of the network.  This technique works
quite well at several sites where I was brought in to diagnose / solve high 
local traffic problems found to be caused by diskless workstations.  

snoopy@sopwith.UUCP (Snoopy) (07/12/89)

In article <2467@internal.Apple.COM> fair@apple.com (Erik E. Fair) writes:

|Just try getting Ultrix source from DEC. When you do, you'll find out that 
|it is "for reference purposes only."

|        Erik E. Fair      apple!fair      fair@apple.com

Presumably this implies that Apple supplies full source code for all its
offerings.

    _____     						  .-----.
   /_____\    Snoopy					./  RIP	 \.
  /_______\   qiclab!sopwith!snoopy			|  	  |
    |___|     parsely!sopwith!snoopy			| tekecs  |
    |___|     sun!nosun!illian!sopwith!snoopy		|_________|

	    "But we're only up to the fourth inning."