That would be cool if no outside programming was required - especially
in light of Ben's comments regarding small image transfers (sorry, I'm
always thinking "on the large"  :^)

stephen

Brian Messenger wrote:
> 
> My solution was one which might work with out any programmer intervention.
> I will delve into the answer further when I have a small cluster to set up
> test and play with.  I was mostly interested to get the input as to whether
> this is a good/bad idea.
> 
> Brian Messenger
> Senior Sales Engineer
> RackSaver, Inc.
> (858)874-3800 x160
> 
> -----Original Message-----
> From: [EMAIL PROTECTED]
> [mailto:[EMAIL PROTECTED]]On Behalf Of Stephen L.
> Scott
> Sent: Tuesday, December 10, 2002 7:33 PM
> To: Brian Messenger
> Cc: [EMAIL PROTECTED]
> Subject: Re: [Oscar-users] Question.
> 
> hi Brian
> 
> off the cuff...
> 
> i don't think it will scale as perfectly as you may think.  the mac
> address acquisition process will finish well before the images are
> pumped out to the nodes. thus causing the same image server bottle-neck
> that we currently have when trying to build large numbers of nodes.
> 
> however, there is a sunny side to your proposition... (this may be what
> you were saying?)  we could follow the parallel programming model of
> "bag of jobs" only in our case it would be a "bag of nodes to build".
> we set a limit as to the number of concurrent builders permitted and as
> one finishes another is queued up to the task...  while this will not
> significantly improve performance, it takes the human interaction out of
> the loop.  would be a reasonable interim solution until something better
> comes along.
> 
> how attractive to do?  that depends on the effort in making it happen
> and the amount of resources (programmers) taken away from developing a
> great new method to work on this interim technique.
> 
> basically we all need a significantly faster build process that can
> handle 1000's of nodes simultaneous.
> 
> stephen
> 
> Brian Messenger wrote:
> >
> > While collecting mac addresses is it possible to image the nodes?
> >
> > My thought was that I could set up a simple dhcpd.conf, click the
> configure
> > PXE server button, next click start collecting button.  The first node I
> > turn on will get the first IP in my spool, I will also get the mac address
> > for the new dhcpd.conf.  As long as nothing else takes any of the
> addresses
> > in my spool this should work fine.  no?
> >
> > This would ultimately eliminate one step in the process of setting up
> oscar
> > for me.  The mac address collection process scales perfectly so the more
> > nodes you add the more time it takes.  Dose anyone have any ideas on this
> as
> > far as, 'would this actually reduce my time installing oscar?'
> >
> > Thank you for any insight.
> >
> > Brian Messenger
> >
> > -------------------------------------------------------
> > This sf.net email is sponsored by:
> > With Great Power, Comes Great Responsibility
> > Learn to use your power at OSDN's High Performance Computing Channel
> > http://hpc.devchannel.org/
> > _______________________________________________
> > Oscar-users mailing list
> > [EMAIL PROTECTED]
> > https://lists.sourceforge.net/lists/listinfo/oscar-users
> 
> --
> ------------------------------------------------------------------------
> Stephen L. Scott, Ph.D.                 voice: 865-574-3144
> Oak Ridge National Laboratory           fax:   865-574-0680
> P. O. Box 2008, Bldg. 6012, MS-6367     [EMAIL PROTECTED]
> Oak Ridge, TN 37831-6367                http://www.csm.ornl.gov/~sscott/
> ------------------------------------------------------------------------
> 
> -------------------------------------------------------
> This sf.net email is sponsored by:
> With Great Power, Comes Great Responsibility
> Learn to use your power at OSDN's High Performance Computing Channel
> http://hpc.devchannel.org/
> _______________________________________________
> Oscar-users mailing list
> [EMAIL PROTECTED]
> https://lists.sourceforge.net/lists/listinfo/oscar-users
> 
> -------------------------------------------------------
> This sf.net email is sponsored by:
> With Great Power, Comes Great Responsibility
> Learn to use your power at OSDN's High Performance Computing Channel
> http://hpc.devchannel.org/
> _______________________________________________
> Oscar-users mailing list
> [EMAIL PROTECTED]
> https://lists.sourceforge.net/lists/listinfo/oscar-users

-- 
------------------------------------------------------------------------
Stephen L. Scott, Ph.D.                 voice: 865-574-3144
Oak Ridge National Laboratory           fax:   865-574-0680
P. O. Box 2008, Bldg. 6012, MS-6367     [EMAIL PROTECTED]
Oak Ridge, TN 37831-6367                http://www.csm.ornl.gov/~sscott/
------------------------------------------------------------------------


-------------------------------------------------------
This sf.net email is sponsored by:
With Great Power, Comes Great Responsibility 
Learn to use your power at OSDN's High Performance Computing Channel
http://hpc.devchannel.org/
_______________________________________________
Oscar-users mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/oscar-users

Reply via email to