*BSD News Article 36712


Return to BSD News archive

Newsgroups: comp.os.386bsd.misc
Path: sserve!newshost.anu.edu.au!harbinger.cc.monash.edu.au!bunyip.cc.uq.oz.au!munnari.oz.au!spool.mu.edu!howland.reston.ans.net!usc!nic-nac.CSU.net!charnel.ecst.csuchico.edu!csusac!csus.edu!netcom.com!NewsWatcher!user
From: tgray@netcom.com (Tom Gray)
Subject: Re: dual procesor motherboards the way forward?
Message-ID: <tgray-0710940002460001@192.187.228.31>
Sender: netnews@netcom.com (USENET Administration)
Organization: August Software
References: <Cww6x9.1A0@gnome.co.uk>
Date: Fri, 7 Oct 1994 08:02:46 GMT
Lines: 78

In article <Cww6x9.1A0@gnome.co.uk>, jacs@gnome.co.uk (Dr Chris Stenton ) wrote:

> I have just had some information sent to me on a dual pentium PCI-EISA
> motherboard which is now availablle.

Yes, in the next year or so, you will be seeing many dual- and multi-
processor motherboards being introduced because of Windows NT.  They are
likely to be very expensive, though.

Wouldn't it would be great if someone produced an inexpensive
mass-merchandised motherboard with between six and eight 40MHz 386DX clone
chips installed?  With the right (SMP aware) operating system, a $500
motherboard like that would scream.

> The blurb says that there is some crude (my word) support under windows NT
> where you can select an application to run on the second processor.

No, Windows NT supports full SMP (Symetrical Multi-Processing).  Not only
are the applications automatically scheduled across available processors,
but so is the operating system itself.  Further, the applications
themselves are broken down into threads which are split accross
processors.  The great thing about all of this is that it is really very
inexpensive (<$400).

There are some UNIX software companies out there that have toolkits
available now for NT that make NT's shells, utilities, and API's "feel"
like UNIX.  There is also at least one company out there with a product to
run a UNIX kernel on top of NT.  (Now, if they can just do something about
surplanting the Windows "look" with Motif or OpenStep :)

Anyway, NT is (politically) decended from (DEC) VAX/VMS, which has had
very, very advanced SMP support since version 5.  (DEC was a pioneer in
multi-processing systems, going way back to their VAX-11/782.)

> This made think what use this additional technology could have for intel based
> unix.

At SUNWorld '94, Scott McNealy (SUN Microsystems) made a big deal out of
SUN's  efforts to bring SMP to their Solaris products.  Solaris/Intel is a
little pricey though.  I'd sure like to see it in the same price-range as
Windows NT, but I think there is an economy-of-scale problem there.

> Would it be better to just place the kernel on one processor and run all other
> processes on the other

This is called ASymetrical Multi-Processing (ASMP) and, although it much
easier to implement and program for, it does not offer as much performance
as SMP for non-compute intensive applications.  This is because most
typical applications spend most of their time in the O/S (doing windowing,
networking, I/O, memory management, etc..), which makes the "root"
processor a resource bottleneck.  Still, for certain types of applications
(simulations, etc..) it does a very good job.

>; or some form of dynamic load balance which will have some overhead?

Right, another route would be to tie physically separate computer systems
together by their motherboard-busses (using 100MB/s network cards) and use
DCE aware applications which can spread their threads evenly over the
available systems.  The issue, of course, is that it is hard to find DCE
aware applications.  (Of course, if DCE is ported to *BSD and Linux, that
could change overnight.)

>Anyway I am sure the FreeBSD and NetBSD teams
> could have some fun with this after they have tired with porting 4.4
> BSD.

It would sure distinguish *BSD from Linux.  Clearly it is only a matter of
time before the Linux anarchy hacks a simple ASMP kernel mod (once
multi-CPU motherboards become more prevelant).  The structure of the *BSD
teams might enable them to go the distance though and implement a true SMP
kernel.  (It would be much more difficult to implement an SMP kernel in an
anarchistic development environment.)  Even still, it could take at least
a work-year to implement SMP.  Perhaps just adding DCE support to the *BSD
kernels would be more realistic.

In the mean-time, there is a version of Linux available (from DEC Western
Research Labs) as well as OSF-1 for DEC's Alpha series, the fastest
processors on the planet.