Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 28 Jul 1998 11:29:54 -0400 (EDT)
From:      Daniel Frasnelli <dfrasnel@csee.wvu.edu>
To:        dbader@eece.unm.edu (David A. Bader)
Cc:        ports@FreeBSD.ORG
Subject:   Re: Ports category submission (fwd)
Message-ID:  <199807281530.LAA20430@naur.cs.wvu.edu>
In-Reply-To: <199807280151.TAA16371@jalapeno.eece.unm.edu> from "David A. Bader" at "Jul 27, 98 07:51:44 pm"

next in thread | previous in thread | raw e-mail | index | archive | help
Greetings,
	
> I think this is a great idea!  I'm the maintainer of the "MPI" port
> (mpich), and would volunteer to help out in any way with developing
> this category. 
	That's wonderful, glad to hear that more folks are interested. 
> The LinuxFolks have a similar distribution called ExtremeLinux which is a 
> descendent of the Beowulf project; and includes all of the tools necessary 
> to make a parallel cluster. 
	Herein lay a "sticky point" which I may or may not catch flak for
bringing up.  The folks at the Beowulf project (and now ExtremeLinux) are 
building computational clusters, some which include facilities for DSM.  
Personally, I define a true "parallel cluster" as something which provides 
not only shared processor power, but also shared memory, a distributed
filesystem, and (to some degree) a shared userspace.  
	Granted, I haven't looked at the Beowulf documentation lately and they
may have these facilities available now.  One fellow I spoke to agreed with
me that a true parallel cluster should appear as one system, in all aspects, 
to the developer or user.  Many freely-available extensions for MPI/PVM exist 
to provide facilities such as a DSM (Adsmith), distributed filesystem
(CODA provides an excellent base), virtual framebuffer, parallel I/O (ROMIO),
etc.  
	Providing a "virtual parallel environment" could possibly be achieved
by implementing a subset of the Mach microkernel as a daemon or set of daemons
whose peripherals are highly abstract and rely on external (network-bound)
nodes for all services.  In this sense, there could concurrently exist both a 
parallelized and "normal" layer of the operating system; userland could be a
minimal subset of the monolithic system shared across a DFS. 
> As a category, I vote for any of: "hpc" (high performance computing), 
> "parallel" (parallel processing), or "cluster" (cluster computing). 
	I second your nominations.  Why don't we go all out and call it
"wolfpack", just to ruffle Microsoft's feathers a bit (Just kidding :-)

Best regards,
Daniel
---
Daniel J. Frasnelli             Imaging spectroscopy research
(dfrasnel@wvu.edu)              Remotely sensed data analyst
Ecologist                       Extreme backpacker


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-ports" in the body of the message



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199807281530.LAA20430>