Re: WAY WAY OFFTOPIC: Message passing vs. shared memory (WAS: Re

Thomas Heide Clausen (voop@innocent.com)
Sun, 25 Jan 1998 12:53:15 +0100 (CET)


-----BEGIN PGP SIGNED MESSAGE-----

On 25-Jan-98 Stephen Williams wrote:
>
> To throw fuel on the fire... :-)
>
> voop@innocent.com said:
>> Which is better or worse of message passing or shared memory
>> is a
>> matter of religion: the professors usually beat eachother up
>> with
>> arguments for which is of preference.
>
> A simple existence proof can show that message passing is
> computationally
> equivalent to shared memory w/ semaphores. algorithms tuned
> for each can
> be implemented with the other.

But I agree completely....they can be shown to be equivalent on
the computational level. However in the world of performance
(Performance in the interpretation of speed and ease of
handeling) they have very different characteristics.....as you
mention yourself below....

As for the equivalence as a reason for stopping a holy war: I
think I mentioned the discussions between synchronous and
asynchronous communication......that hasn't stopped yet either
(and most likely will never do).....

>
> I've done very large distributed systems and find that message
> passing
> better reflects the distributed reality of large systems. When
> you start
> getting into error handling and faults, that memory doesn't
> normally have,
> you start to realize why communication protocols are described
> with messages.
>

There you go.....you find message passing of preference in some
situations - exactly that kind of arguments I was refering
to.....

My own arguments for not using distributed shared memory but
message passing has something to to with independance of the
latency between the nodes (thus the memory) in a distributed
system. (which is allso a reason for using asynchronous
messages). But let us not take this much further - I still think
that pro/con for shared memory and/or message passing is WAY WAY
off topic.

> DSHM is a neat idea, and should be provided by Linux,
> especially for the
> smaller problems, but the real nasty systems are going to be
> using some
> sort of message passing protocol.

Sure...Let's have everything - I agree completely.

>
> --
> Steve Williams "The woods are lovely, dark and
> deep.
> steve@icarus.com But I have promises to keep,
> steve@picturel.com and lines to code before I
> sleep,
> http://www.picturel.com And lines to code before I
> sleep."

- --thomas

- ---------------------------------------------------
E-Mail: Thomas Heide Clausen <voop@innocent.com>
URL: http://www.cs.auc.dk/~voop/

Masters student in the distributed systems group
Department for Computer Science, Aalborg University

Visitor at University of Illinois
Urbana-Champaign

Date: 25-Jan-98, Time: 12:39:12

Unite for Java! - http://www.javalobby.org
- ---------------------------------------------------

-----BEGIN PGP SIGNATURE-----
Version: 2.6.3ia
Charset: noconv

iQCVAwUBNMsnOcQLb2bL5bWVAQFcvAP+PMNPfKV8x7jtJVtzOAT/N+topCAGaXt0
aFn/FMv9l/rq8UfsmFIi7zQiIxlZf6dfj2Pxt6YTVjnRJeyb4qiC5q3yrN+Q2fm0
VKRYlu5lsygBwERpTwiNdVbxxVEfZkrLEoIf2E27Yt0H/PSDpFD9Y7TDxdmwdIAc
6iIKIlQirbg=
=FKbK
-----END PGP SIGNATURE-----