Newsgroups: comp.parallel.pvm
From: crispin@csd.uwo.ca (Crispin Cowan)
Subject: Re: PVM under Windows?
Organization: Department of Computer Science, University of Western Ontario, London, Ontario, Canada
Date: 24 Aug 1994 20:27:28 GMT
Message-ID: <33gafg$o4g@falcon.ccs.uwo.ca>

In article <Y0VRBY4T@math.fu-berlin.de>,
Thomas Wolff <wolff@inf.fu-berlin.de> wrote:
>The apparently prevailing purpose of PVM is to provide easier means of 
>communication between processes running on different machines - this is 
>distribution, despite the fact that it uses lower-level systems that 
>also provide distribution.
The parallel processing community and the distributed processing
community are largely disjoint.  Despite the fact that they both use
similar concurrency models, primitives, and sometimes even hardware,
where the differ is intent.

In parallel processing, the intent is to use multiple machines to get a
given job done faster.  In distributed processing, the intent is to get
diverse, autonomous machines to co-operate in order to advance each of
their own goals, which sometimes leads to parallel processing.

>: It's in the name:  PVM == Parallel Virtual Machine.
>This is not an argument.
It goes to intent.  PVM is intended to support parallel processing, and
is not intended to support distributed processing.  There is no support
for autonomy in PVM what so ever:  the entire virtual machine is
controlled from the console.  While the console can be launched on any
of the nodes in the virtual machine, there had better be only one at
any given time.

>: PVM provides the primitives one expects to see in a
>: MPP environment (message passing, array packing, etc.).
>Now this is definately distribution, not parallelism.
Nope.  MPP (Massively Parallel Processors) are machines consisting of a
large number of processing elements in a box.  Sometimes MPPs are
shared memory machines, and sometimes they are message passing machines
(such as the Intel Paragon, the nCUBE, and the Cray T3D).  They are not
distributed machines.

>: environment:  resource brokers, multi-user capabilities, security,
>: authentication, the ability for an outside task to add itself to the
>: system, etc.
>These are advanced features not necessarily inherent to a simple 
>distributed system. It is one of the very appreciable features that 
>PVM lacks much of this fuzz so it can be employed easier for 
>systems with simpler requirements.

PVM lacks these features because it is not intended to be a distributed
system, it is intended to emulate the above mentioned parallel systems
in a distributed environment.  Hence the name:  Parallel Virtual
Machine, i.e. it emulates a single parallel machine.

Crispin
-----
Crispin Cowan, CS PhD student, searching for a research position
University of Western Ontario
Phyz-mail:  Middlesex College, MC28-C, London, Ontario, N6A 5B7
E-mail:     crispin@csd.uwo.ca          Voice:  519-661-3342
		Stop Photo RADAR:  Fight your ticket!

