Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!wuarchive!udel!rochester!pt.cs.cmu.edu!gandalf.cs.cmu.edu!lindsay
From: lindsay@gandalf.cs.cmu.edu (Donald Lindsay)
Newsgroups: comp.arch
Subject: Shared Memory (was Terradata architecures)
Message-ID: <10651@pt.cs.cmu.edu>
Date: 4 Oct 90 01:55:25 GMT
References: <211@bilpin.UUCP> <1990Sep28.020717.22610@dhw68k.cts.com> <1990Oct1.200613.635@tera.com> <1990Oct3.013509.1470@news.iastate.edu>
Organization: Carnegie-Mellon University, CS/RI
Lines: 45

In article <1990Oct3.013509.1470@news.iastate.edu> carter@iastate.edu 
	(Carter Michael Brannon) writes:
>While I must agree
>with you that shared-memory systems are far from dinosaurs, I must disagree
>categorically that they are the only cost-effective way to build a successful
>massively-parallel computer.

It is clear that it is simplest to design a message-based ensemble
machine.  A lot of issues can be punted to software, and the hardware
that is designed, pretty well has to be there in any case.

It is also clear that a shared-memory ensemble machine is desirable.
Sharing allows some nice programming techniques, and makes a
difference when trying to port a lot of software to the new machine.

So: the shared-memory machine is harder to design, but (broadly) more
valuable.  Now for the Big Question: is it necessarily more
expensive?

I don't think so, although that's easier to say than prove.

To begin with, adding function to an existing chip may be free, from
the manufacturing viewpoint.  This idea hints to us that increasing a
design's complexity doesn't necessarily increase its price.

Next, at what level will we share?  If it's at the level of virtual
memory pages, then this is already being done, over LANs, by several
operating systems (e.g. Mach).  The only hardware facilities required
are 1) a decent MMU on each node, and 2) page-sized messages.

The problem with many current ensemble machines is their indecent
MMUs.  I submit that an MMU is valuable in any case, so page-sharing
makes no special hardware demands: zero, none.  Not that hardware
couldn't be useful: and the interconnect has to be fast enough so
that the machine remains balanced under the new usage patterns.

If a machine is to have finer-grained sharing, then a number of
efficiency issues come up.  I am aware of a variety of approaches,
each with its own penalty or price.  I believe that eventually, one
of these will be implemented in a way that adds essentially nothing
to the machine's manufacturing cost.  


-- 
Don		D.C.Lindsay