<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=Windows-1252">
<META NAME="Generator" CONTENT="MS Exchange Server version 6.0.6619.9">
<TITLE>RE: [Beowulf] Win64 Clusters!!!!!!!!!!!!</TITLE>
</HEAD>
<BODY>
<!-- Converted from text/plain format -->
<P><FONT SIZE=2>Thanks John,<BR>
<BR>
I appreciate your response. It is what I thought but I wasn't quite sure. I looked at MPICH2 for Windows and it does have two techniques with tradeoffs as you mentioned...I hope that you do decide to open up things a little more and allow other job schedulers and MPI stacks access to the credential management. I know from experience that the current job schedulers on Windows (other than your own) do not handle credentials in a transparent manner and it would be nice if their credential management was better implemented....entering a user password and saving it...'somewhere safe' is not really an ideal solution especially when the users password changes and the job scheduler does not pickup the change.<BR>
<BR>
Regards,<BR>
<BR>
Bill Bryce<BR>
Product Manager<BR>
Platform Open Cluster Stack<BR>
<BR>
<BR>
<BR>
-----Original Message-----<BR>
From: John Vert [<A HREF="mailto:jvert@windows.microsoft.com">mailto:jvert@windows.microsoft.com</A>]<BR>
Sent: Tue 4/3/2007 10:32 PM<BR>
To: Bill Bryce; Robert G. Brown<BR>
Cc: beowulf@beowulf.org<BR>
Subject: RE: [Beowulf] Win64 Clusters!!!!!!!!!!!!<BR>
<BR>
The job scheduler is responsible for managing the user credentials and<BR>
tokens as well as allocating CPUs to jobs. So if you are not going to<BR>
use our scheduler, you need a different solution for the credential<BR>
management. How do the user's processes get created with the right<BR>
token? MPICH2 has a couple different solutions with different tradeoffs<BR>
- I'm not sure which one you use.<BR>
<BR>
We have had requests for two different scenarios - enabling other MPI<BR>
stacks to use the credential management of our job scheduler and<BR>
enabling other job schedulers to do the credential management for our<BR>
MPI stack. We are looking at both of these for the next release but they<BR>
didn't make V1.<BR>
<BR>
So like most product decisions, we didn't "have to do this" it was just<BR>
the normal tradeoff you make between features and schedules with a<BR>
healthy dose of security paranoia thrown into the mix.<BR>
<BR>
John Vert<BR>
Development Manager<BR>
High Performance Computing<BR>
<BR>
-----Original Message-----<BR>
From: Bill Bryce [<A HREF="mailto:bill@platform.com">mailto:bill@platform.com</A>]<BR>
Sent: Tuesday, April 03, 2007 1:12 PM<BR>
To: John Vert; Robert G. Brown<BR>
Cc: beowulf@beowulf.org<BR>
Subject: RE: [Beowulf] Win64 Clusters!!!!!!!!!!!!<BR>
<BR>
Can I ask one simple question.....<BR>
<BR>
Why did Microsoft have to do this (and I quote from MSDN)<BR>
<BR>
"Every job and task that uses Microsoft MPI must be submitted through<BR>
the Microsoft job scheduler. There are several ways to submit jobs and<BR>
tasks to the job scheduler:"<BR>
<BR>
I do not have that restriction with other MPI's such as MPICH2 for<BR>
Windows from Argonne.<BR>
<BR>
Regards,<BR>
<BR>
Bill.<BR>
<BR>
<BR>
-----Original Message-----<BR>
From: John Vert [<A HREF="mailto:jvert@windows.microsoft.com">mailto:jvert@windows.microsoft.com</A>]<BR>
Sent: Tuesday, April 03, 2007 3:19 PM<BR>
To: Robert G. Brown; Bill Bryce<BR>
Cc: beowulf@beowulf.org<BR>
Subject: RE: [Beowulf] Win64 Clusters!!!!!!!!!!!!<BR>
<BR>
I'd like to clear up a few misconceptions on this thread.<BR>
<BR>
1. There is no "MS specific library that abuses 'standard MPI'". The #1<BR>
goal for our MPI implementation is to make it as simple as possible for<BR>
software vendors to port their codes to Windows and our MPI stack. That<BR>
is why we based our implementation on MPICH2. The feedback we have<BR>
gotten from software vendors has shown this approach was successful and<BR>
we have had very few issues with MPI incompatibilities. We have not<BR>
added anything outside the MPI spec.<BR>
2. MPICH2 is not licensed under GPL or under BSD. See<BR>
<A HREF="http://www-unix.mcs.anl.gov/mpi/mpich2/license.htm">http://www-unix.mcs.anl.gov/mpi/mpich2/license.htm</A> Many commercial<BR>
closed-source MPI implementations have been based on MPICH. Microsoft is<BR>
actively working with Argonne on incorporating the changes we made back<BR>
into the MPICH2 distribution. Argonne has been great to work with and<BR>
really supportive of our efforts.<BR>
3. The changes we made can be loosely grouped into three areas.<BR>
Performance, security, and scheduler integration. We handle the user<BR>
credentials differently than Argonne's Windows implementation which is<BR>
different than Argonne's Linux implementation due to some fundamental<BR>
differences between the two operating systems. The scheduler integration<BR>
is pretty straightforward - things like cleaning up all MPI child<BR>
processes when a job is cancelled, and automatically propagating things<BR>
like the allocated nodes and the number of CPUs from the scheduler to<BR>
mpiexec. I don't think this is significantly different than what other<BR>
job schedulers have done.<BR>
4. If you want to learn more about Windows HPC clusters, I recommend<BR>
checking out www.microsoft.com/hpc and www.windowshpc.net. I'm also<BR>
happy to answer questions on this list, but frankly the S/N ratio tends<BR>
to drop dramatically as soon as someone mentions Windows or Microsoft.<BR>
<BR>
John Vert<BR>
Development Manager<BR>
High Performance Computing<BR>
<BR>
-----Original Message-----<BR>
From: beowulf-bounces@beowulf.org [<A HREF="mailto:beowulf-bounces@beowulf.org">mailto:beowulf-bounces@beowulf.org</A>]<BR>
On Behalf Of Robert G. Brown<BR>
Sent: Tuesday, April 03, 2007 7:59 AM<BR>
To: Bill Bryce<BR>
Cc: beowulf@beowulf.org<BR>
Subject: RE: [Beowulf] Win64 Clusters!!!!!!!!!!!!<BR>
<BR>
On Tue, 3 Apr 2007, Bill Bryce wrote:<BR>
<BR>
> So this change effectively ties Microsoft MPI to only Microsoft<BR>
Windows<BR>
> platforms, and the security changes are closed source. Not all of<BR>
> Microsoft's partners like MS MPI - when HP ships Microsoft CCS they<BR>
> remove MS MPI and put in HP MPI - which probably just adds to<BR>
confusion<BR>
> on the end user side.<BR>
<BR>
Security changes? To MPI itself? This puts them into a somewhat grey<BR>
area with regard to the GPL, doesn't it... that viral thing.<BR>
<BR>
I'm sure that they can manage the remote job launch any way that they<BR>
like, but doesn't that leave the MPI CODE still portable?<BR>
<BR>
It isn't completely crazy that the US government would intervene here if<BR>
they broke MPI portability. After all, MPI exists at all primarily<BR>
because of direct government intervention (unlike PVM, which exists<BR>
because some Very Bright People conceived it and invented it and made it<BR>
fly).<BR>
<BR>
Still, nothing the Borg Empire does in this regard would surprise me.<BR>
<BR>
rgb<BR>
<BR>
><BR>
> Regards,<BR>
><BR>
> Bill.<BR>
><BR>
><BR>
> -----Original Message-----<BR>
> From: beowulf-bounces@beowulf.org [<A HREF="mailto:beowulf-bounces@beowulf.org">mailto:beowulf-bounces@beowulf.org</A>]<BR>
> On Behalf Of Robert G. Brown<BR>
> Sent: Monday, April 02, 2007 8:22 PM<BR>
> To: Tom Mitchell<BR>
> Cc: beowulf@beowulf.org<BR>
> Subject: Re: [Beowulf] Win64 Clusters!!!!!!!!!!!!<BR>
><BR>
> On Mon, 2 Apr 2007, Tom Mitchell wrote:<BR>
><BR>
>> On Sun, Apr 01, 2007 at 02:02:53PM +0500, amjad ali wrote:<BR>
>>> Date: Sun, 1 Apr 2007 14:02:53 +0500<BR>
>>> From: "amjad ali" <amjad11@gmail.com><BR>
>>> To: beowulf@beowulf.org<BR>
>>> Subject: [Beowulf] Win64 Clusters!!!!!!!!!!!!<BR>
>>><BR>
>>> Hi All,<BR>
>>><BR>
>>> Would any of you please like to share<BR>
usage-experience/views/comments<BR>
>>> about Windows Compute Cluster Server 2003 based Beowulf Clusters?<BR>
>>><BR>
>>> What in your opinion is the future of such clusters?<BR>
>>><BR>
>>> How you compare these with the LINUX CLUSTERS?<BR>
>><BR>
>> With full consideration to "fat, soft pretzels with<BR>
>> cheezy-mustard sauce or rolled in asiago parmesan and garlic."<BR>
>><BR>
>> MS pulled a version of mpich/mvapich/MPI and ported it to windows.<BR>
>> They also developed some library code to gateway some *inx<BR>
> library/system<BR>
>> calls to windows. The root sources of MPI are public and not GPL so<BR>
> they can.<BR>
>><BR>
>> It might be worth looking at the MS announcement -- but why<BR>
>> bother. If you look you might think that common MPI codes<BR>
>> would just compile and run... I have no idea I expect some will<BR>
>> and there begins silly porting for the next...<BR>
><BR>
> Sure. MS did this, no doubt. And as you note below, no sooner do<BR>
they<BR>
> get it in when they begin the borgification of MPI, just as they've<BR>
> borgified java, c, c++, and anything in the Universe they can sucker<BR>
> somebody into buying in borgified form.<BR>
><BR>
> Borgifying MPI is the most humorous thing in the Universe, BTW, given<BR>
> its historical origins -- it was basically a language written<BR>
> (reluctantly!) by supercomputer vendors when the US government got<BR>
tired<BR>
> of paying for all their important codes to be ported to each new<BR>
> generation of proprietary hardware with its proprietary low level<BR>
calls.<BR>
> MS is doubtless trying to figure out just how much of that they can<BR>
> undo while building up a big enough market share and enough vendors of<BR>
> closed source applications written with their borgisms that they<BR>
can...<BR>
><BR>
> Oh wait. It IS GPL. Do you think that they actually read it?<BR>
><BR>
> However, I was really referring to the other aspects of program<BR>
> development and performance tuning associated with using a closed<BR>
source<BR>
> development environment.<BR>
><BR>
> Resistance is Futile.<BR>
><BR>
> rgb<BR>
><BR>
>><BR>
>> Once a set of boxes are interconnected and you have library<BR>
>> support for MPI or another way to share data (PVM... whatever)<BR>
>> you are off and running in the clustering world. Sadly MS<BR>
>> has a MS specific library that abuses "standard MPI" and could<BR>
>> quickly cause source code to surface that runs correctly or on a<BR>
>> MS cluster but not on another OS based cluster (Linux, Solaris,<BR>
>> Irix, AIX). I see this all the time with java script, and c,<BR>
>> c++, and other codes where little 'features' hook you in.<BR>
>><BR>
>> Some will be fooled into thinking that this is something to look at<BR>
>> or worse something to spend money on.<BR>
>><BR>
>> SUMMARY:<BR>
>> Since you posted this on 1 Apr 2007 all I can do is giggle<BR>
>> and wonder why I replied.<BR>
>><BR>
>> Regards,<BR>
>> mitch<BR>
>><BR>
>> PS: Ask in a year but not on April fools/joke day.<BR>
>><BR>
>><BR>
><BR>
><BR>
<BR>
--<BR>
Robert G. Brown <A HREF="http://www.phy.duke.edu/~rgb/">http://www.phy.duke.edu/~rgb/</A><BR>
Duke University Dept. of Physics, Box 90305<BR>
Durham, N.C. 27708-0305<BR>
Phone: 1-919-660-2567 Fax: 919-660-2525 email:rgb@phy.duke.edu<BR>
<BR>
<BR>
_______________________________________________<BR>
Beowulf mailing list, Beowulf@beowulf.org<BR>
To change your subscription (digest mode or unsubscribe) visit<BR>
<A HREF="http://www.beowulf.org/mailman/listinfo/beowulf">http://www.beowulf.org/mailman/listinfo/beowulf</A><BR>
<BR>
<BR>
<BR>
</FONT>
</P>
</BODY>
</HTML>