[AG-TECH] Fwd: BOUNCE ag-tech at mcs.anl.gov: Non-member submission from [Douglas Baggett <dbaggett at nsf.gov>]

[Justin Binns]

My understanding of MOSIX is that it accomplishes clustering by
distributing the user-space portion of a process, while making all system
calls back to the originating machine over the network interconnect.  If
that is indeed the case, than running RAT/VIC over MOSIX would probably
*decrease* performance considerably (every network send/receive is a
write/read call, which crosses the user/kernel space boundary, and would
therefore be run indirectly).

Justin

On Wed, 16 Oct 2002, Robert Olson wrote:

> 
> >Date: Tue, 15 Oct 2002 07:34:19 -0500
> >From: owner-ag-tech at mcs.anl.gov
> >To: owner-ag-tech at mcs.anl.gov
> >Subject: BOUNCE ag-tech at mcs.anl.gov:    Non-member submission from 
> >[Douglas Baggett <dbaggett at nsf.gov>]
> >X-Spam-Status: No, hits=2.0 required=5.0 
> >tests=NO_REAL_NAME,X_NOT_PRESENT,FROM_AND_TO_SAME version=2.21
> >X-Spam-Level: **
> >
> >Received: from lapp1.cise-nsf.gov (lapp1.cise-nsf.gov [192.12.209.21])
> >         by mcs.anl.gov (8.9.3/8.9.3) with ESMTP id HAA119072
> >         for <ag-tech at mcs.anl.gov>; Tue, 15 Oct 2002 07:34:16 -0500
> >Received: from db-cise-worksta.nsf.gov (db-cise-worksta.cise-nsf.gov 
> >[192.12.209.224])
> >         by lapp1.cise-nsf.gov (8.11.6/8.11.6) with ESMTP id g9FCYIQ11513;
> >         Tue, 15 Oct 2002 08:34:18 -0400
> >Message-Id: <5.1.0.14.0.20021015083007.028d55b8 at cisebdc.cise-nsf.gov>
> >X-Sender: dbaggett at cisebdc.cise-nsf.gov
> >X-Mailer: QUALCOMM Windows Eudora Version 5.1
> >Date: Tue, 15 Oct 2002 08:34:16 -0400
> >To: ag-tech at mcs.anl.gov
> >From: Douglas Baggett <dbaggett at nsf.gov>
> >Subject: VIC - OpenMASH or UCL and OpenMOSIX
> >Cc: dgatchel at cise-nsf.gov, rhilderb at nsf.gov
> >Mime-Version: 1.0
> >Content-Type: text/plain; charset="us-ascii"; format=flowed
> >X-Spam-Status: No, hits=1.2 required=5.0 tests=MSG_ID_ADDED_BY_MTA_3 
> >version=2.21
> >X-Spam-Level: *
> >
> >Has anybody tried to run VIC and RAT stand-alone on a LINUX system part of
> >an OpenMOSIX cluster? I don't know enough of how OpenMASH/UCL VIC and RAT
> >are coded to know if one would see any performance gain from being part of
> >a cluster ( I also am pretty much a newbie when it comes to working with
> >clustering).
> >
> >I was thinking it might be neat to try it out.
> >
> >Does anybody have any comments on if it could be worth it (other than the
> >pure cool factor involved...)
> >
> >-Doug Bagett
> >CISE/NSF/OAD
> 
>