March 2008 - mpiwg-ft - mailman3.cels.anl.gov

Re: [Mpi3-ft] Summary of today's call
by Supalov, Alexander 31 Mar '08

31 Mar '08

Dear Edgar, Thank you. I've not thought about error classes for fault handling yet, but most likely they will have to be added. The files, etc. are mentioned in the second half of the proposal (I hope). In any case, the intention has always been to use any of the objects that can carry error handlers for fault tolerance as well. Best regards. Alexander -----Original Message----- From: mpi3-ft-bounces(a)lists.mpi-forum.org [mailto:[email protected]] On Behalf Of Edgar Gabriel Sent: Monday, March 31, 2008 6:52 PM To: MPI 3.0 Fault Tolerance and Dynamic Process Control working Group Subject: Re: [Mpi3-ft] Summary of today's call Alexander, the section on fault-handling on the wiki does not mention error classes. Do we want to address that, i.e. does the concept make sense and we suggest to keep it, do we modify the concept since the original idea and its benefits are not entirely clear or do we want to get rid of error classes? second, if I remember right, MPI_FILE_NULL has a different default error handler than MPI_COMM_WORLD. Do we want to mention somewhere in the document, that different objects might have different default error handlers attached? You have a distinction between COMM_WORLD and COMM_SELF -- which is good, but maybe we should explicitly mention other objects as well... Thanks Edgar Supalov, Alexander wrote: > Dear Rich, > > Thanks. We had a scheduled site maintenance and my phone connection was > terminated for some reason. Sorry for that. > > Sure, I can take on fleshing out the rest of the fault handling > proposal. I invite everybody to contribute to the respective Wiki page > in the meantime (see > http://svn.mpi-forum.org/trac/mpi-forum-web/wiki/Fault%20Handling) > and/or send comments to me or this list. > > Best regards. > > Alexander > > -----Original Message----- > From: mpi3-ft-bounces(a)lists.mpi-forum.org > [mailto:[email protected]] On Behalf Of Richard Graham > Sent: Saturday, March 29, 2008 12:29 AM > To: MPI 3.0 Fault Tolerance and Dynamic Process Control working Group > Subject: [Mpi3-ft] Summary of today's call > > Here is a summary of the action items from today's call: > - The consumer/producer scenario is ready for discussion before the > full > forum, once we have nailed down the approach to error handling within > MPI. > The plan is for Edgar to bring this to the full forum for discussion at > the > April meeting. > - The approach to error handling is almost ready to bring to the full > forum. Several items came up in today's call (async notification and > handler stacking) that still need to be fleshed out and fully defined. > Alexander, since you were kicked off the call - would you be willing to > coordinate getting this ready for discussion in the full forum setting > at > the April meeting ? We should discuss this again an the call in 2 weeks > (4/11/2008) and go over the remaining items. > - Josh will work with ? to get the proposal for non-blocking dynamic > process creation ready to go. We should aim to continue discussions on > this > next time, but this may not be ready by the April meeting. If the forum > is > amenable to this proposal, it should be integrated with the rest of > dynamic > process control chapter. > > Also, time permitting we will discuss the data piggy backing proposal on > the > next call. > > What did I forget? > Rich > > _______________________________________________ > mpi3-ft mailing list > mpi3-ft(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-ft > --------------------------------------------------------------------- > Intel GmbH > Dornacher Strasse 1 > 85622 Feldkirchen/Muenchen Germany > Sitz der Gesellschaft: Feldkirchen bei Muenchen > Geschaeftsfuehrer: Douglas Lusk, Peter Gleissner, Hannes Schwaderer > Registergericht: Muenchen HRB 47456 Ust.-IdNr. > VAT Registration No.: DE129385895 > Citibank Frankfurt (BLZ 502 109 00) 600119052 > > This e-mail and any attachments may contain confidential material for > the sole use of the intended recipient(s). Any review or distribution > by others is strictly prohibited. If you are not the intended > recipient, please contact the sender and delete all copies. > > > _______________________________________________ > mpi3-ft mailing list > mpi3-ft(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-ft -- Edgar Gabriel Assistant Professor Parallel Software Technologies Lab http://pstl.cs.uh.edu Department of Computer Science University of Houston Philip G. Hoffman Hall, Room 524 Houston, TX-77204, USA Tel: +1 (713) 743-3857 Fax: +1 (713) 743-3335 _______________________________________________ mpi3-ft mailing list mpi3-ft(a)lists.mpi-forum.org http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-ft --------------------------------------------------------------------- Intel GmbH Dornacher Strasse 1 85622 Feldkirchen/Muenchen Germany Sitz der Gesellschaft: Feldkirchen bei Muenchen Geschaeftsfuehrer: Douglas Lusk, Peter Gleissner, Hannes Schwaderer Registergericht: Muenchen HRB 47456 Ust.-IdNr. VAT Registration No.: DE129385895 Citibank Frankfurt (BLZ 502 109 00) 600119052 This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies.

1 0

Re: [Mpi3-ft] Summary of today's call
by Edgar Gabriel 31 Mar '08

31 Mar '08

Alexander, the section on fault-handling on the wiki does not mention error classes. Do we want to address that, i.e. does the concept make sense and we suggest to keep it, do we modify the concept since the original idea and its benefits are not entirely clear or do we want to get rid of error classes? second, if I remember right, MPI_FILE_NULL has a different default error handler than MPI_COMM_WORLD. Do we want to mention somewhere in the document, that different objects might have different default error handlers attached? You have a distinction between COMM_WORLD and COMM_SELF -- which is good, but maybe we should explicitly mention other objects as well... Thanks Edgar Supalov, Alexander wrote: > Dear Rich, > > Thanks. We had a scheduled site maintenance and my phone connection was > terminated for some reason. Sorry for that. > > Sure, I can take on fleshing out the rest of the fault handling > proposal. I invite everybody to contribute to the respective Wiki page > in the meantime (see > http://svn.mpi-forum.org/trac/mpi-forum-web/wiki/Fault%20Handling) > and/or send comments to me or this list. > > Best regards. > > Alexander > > -----Original Message----- > From: mpi3-ft-bounces(a)lists.mpi-forum.org > [mailto:[email protected]] On Behalf Of Richard Graham > Sent: Saturday, March 29, 2008 12:29 AM > To: MPI 3.0 Fault Tolerance and Dynamic Process Control working Group > Subject: [Mpi3-ft] Summary of today's call > > Here is a summary of the action items from today's call: > - The consumer/producer scenario is ready for discussion before the > full > forum, once we have nailed down the approach to error handling within > MPI. > The plan is for Edgar to bring this to the full forum for discussion at > the > April meeting. > - The approach to error handling is almost ready to bring to the full > forum. Several items came up in today's call (async notification and > handler stacking) that still need to be fleshed out and fully defined. > Alexander, since you were kicked off the call - would you be willing to > coordinate getting this ready for discussion in the full forum setting > at > the April meeting ? We should discuss this again an the call in 2 weeks > (4/11/2008) and go over the remaining items. > - Josh will work with ? to get the proposal for non-blocking dynamic > process creation ready to go. We should aim to continue discussions on > this > next time, but this may not be ready by the April meeting. If the forum > is > amenable to this proposal, it should be integrated with the rest of > dynamic > process control chapter. > > Also, time permitting we will discuss the data piggy backing proposal on > the > next call. > > What did I forget? > Rich > > _______________________________________________ > mpi3-ft mailing list > mpi3-ft(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-ft > --------------------------------------------------------------------- > Intel GmbH > Dornacher Strasse 1 > 85622 Feldkirchen/Muenchen Germany > Sitz der Gesellschaft: Feldkirchen bei Muenchen > Geschaeftsfuehrer: Douglas Lusk, Peter Gleissner, Hannes Schwaderer > Registergericht: Muenchen HRB 47456 Ust.-IdNr. > VAT Registration No.: DE129385895 > Citibank Frankfurt (BLZ 502 109 00) 600119052 > > This e-mail and any attachments may contain confidential material for > the sole use of the intended recipient(s). Any review or distribution > by others is strictly prohibited. If you are not the intended > recipient, please contact the sender and delete all copies. > > > _______________________________________________ > mpi3-ft mailing list > mpi3-ft(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-ft -- Edgar Gabriel Assistant Professor Parallel Software Technologies Lab http://pstl.cs.uh.edu Department of Computer Science University of Houston Philip G. Hoffman Hall, Room 524 Houston, TX-77204, USA Tel: +1 (713) 743-3857 Fax: +1 (713) 743-3335

1 0

Re: [Mpi3-ft] Summary of today's call
by Supalov, Alexander 28 Mar '08

28 Mar '08

Dear Rich, Thanks. We had a scheduled site maintenance and my phone connection was terminated for some reason. Sorry for that. Sure, I can take on fleshing out the rest of the fault handling proposal. I invite everybody to contribute to the respective Wiki page in the meantime (see http://svn.mpi-forum.org/trac/mpi-forum-web/wiki/Fault%20Handling) and/or send comments to me or this list. Best regards. Alexander -----Original Message----- From: mpi3-ft-bounces(a)lists.mpi-forum.org [mailto:[email protected]] On Behalf Of Richard Graham Sent: Saturday, March 29, 2008 12:29 AM To: MPI 3.0 Fault Tolerance and Dynamic Process Control working Group Subject: [Mpi3-ft] Summary of today's call Here is a summary of the action items from today's call: - The consumer/producer scenario is ready for discussion before the full forum, once we have nailed down the approach to error handling within MPI. The plan is for Edgar to bring this to the full forum for discussion at the April meeting. - The approach to error handling is almost ready to bring to the full forum. Several items came up in today's call (async notification and handler stacking) that still need to be fleshed out and fully defined. Alexander, since you were kicked off the call - would you be willing to coordinate getting this ready for discussion in the full forum setting at the April meeting ? We should discuss this again an the call in 2 weeks (4/11/2008) and go over the remaining items. - Josh will work with ? to get the proposal for non-blocking dynamic process creation ready to go. We should aim to continue discussions on this next time, but this may not be ready by the April meeting. If the forum is amenable to this proposal, it should be integrated with the rest of dynamic process control chapter. Also, time permitting we will discuss the data piggy backing proposal on the next call. What did I forget? Rich _______________________________________________ mpi3-ft mailing list mpi3-ft(a)lists.mpi-forum.org http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-ft --------------------------------------------------------------------- Intel GmbH Dornacher Strasse 1 85622 Feldkirchen/Muenchen Germany Sitz der Gesellschaft: Feldkirchen bei Muenchen Geschaeftsfuehrer: Douglas Lusk, Peter Gleissner, Hannes Schwaderer Registergericht: Muenchen HRB 47456 Ust.-IdNr. VAT Registration No.: DE129385895 Citibank Frankfurt (BLZ 502 109 00) 600119052 This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies.

1 0

Re: [Mpi3-ft] Call is still going
by Richard Graham 28 Mar '08

28 Mar '08

Greg, The details on the ft working group web site. In a nut shell, this happens every 2 weeks, but not the week of the MPI forum. Rich On 3/28/08 7:42 PM, "Greg Bronevetsky" <bronevetsky1(a)llnl.gov> wrote: > It seems that I didn't have this call on my schedule. Can someone > forward me the schedule of our calls? > > Greg Bronevetsky > Post-Doctoral Researcher > 1028 Building 451 > Lawrence Livermore National Lab > (925) 424-5756 > bronevetsky1(a)llnl.gov > _______________________________________________ > mpi3-ft mailing list > mpi3-ft(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-ft >

1 0

Re: [Mpi3-ft] Call is still going
by Greg Bronevetsky 28 Mar '08

28 Mar '08

It seems that I didn't have this call on my schedule. Can someone forward me the schedule of our calls? Greg Bronevetsky Post-Doctoral Researcher 1028 Building 451 Lawrence Livermore National Lab (925) 424-5756 bronevetsky1(a)llnl.gov

1 0

[Mpi3-ft] Summary of today's call
by Richard Graham 28 Mar '08

28 Mar '08

Here is a summary of the action items from today's call: - The consumer/producer scenario is ready for discussion before the full forum, once we have nailed down the approach to error handling within MPI. The plan is for Edgar to bring this to the full forum for discussion at the April meeting. - The approach to error handling is almost ready to bring to the full forum. Several items came up in today's call (async notification and handler stacking) that still need to be fleshed out and fully defined. Alexander, since you were kicked off the call - would you be willing to coordinate getting this ready for discussion in the full forum setting at the April meeting ? We should discuss this again an the call in 2 weeks (4/11/2008) and go over the remaining items. - Josh will work with ? to get the proposal for non-blocking dynamic process creation ready to go. We should aim to continue discussions on this next time, but this may not be ready by the April meeting. If the forum is amenable to this proposal, it should be integrated with the rest of dynamic process control chapter. Also, time permitting we will discuss the data piggy backing proposal on the next call. What did I forget? Rich

1 0

[Mpi3-ft] Call is still going
by Richard Graham 28 Mar '08

28 Mar '08

1 0

Re: [Mpi3-ft] Next concall?
by Graham, Richard L. 21 Mar '08

21 Mar '08

Next week ----- Original Message ----- From: mpi3-ft-bounces(a)lists.mpi-forum.org <mpi3-ft-bounces(a)lists.mpi-forum.org> To: mpi3-ft(a)lists.mpi-forum.org <mpi3-ft(a)lists.mpi-forum.org> Sent: Fri Mar 21 11:54:43 2008 Subject: [Mpi3-ft] Next concall? Folks, Are we having our MPI3-FT concall this week? (apologies if this was discussed in the March MPI Forum F2F, but I didn't attend it) Thanx! Kannan _______________________________________________ mpi3-ft mailing list mpi3-ft(a)lists.mpi-forum.org http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-ft

1 0

[Mpi3-ft] Next concall?
by Narasimhan, Kannan 21 Mar '08

21 Mar '08

Folks, Are we having our MPI3-FT concall this week? (apologies if this was discussed in the March MPI Forum F2F, but I didn't attend it) Thanx! Kannan

1 0

Re: [Mpi3-ft] Quiescence Interface
by Greg Bronevetsky 10 Mar '08

10 Mar '08

I just read over the proposal and I still don't see a detailed definition of quiescence. I know what it means when we're talking about checkpointing at the kernel-level(BLCR) or user-level(libckpt) or the application level(C3) and I can figure out what it means when the checkpointer works at the virtual machine monitor level, though that also depends on how well the VMM checkpointer supports the machine's network card. However, what does the spec mean? The draft spec doesn't provide any details and my concern is that the spec can never provide enough details without going into deep details of the network hardware and software stack. Greg Bronevetsky Post-Doctoral Researcher 1028 Building 451 Lawrence Livermore National Lab (925) 424-5756 bronevetsky1(a)llnl.gov At 06:14 PM 3/10/2008, Josh Hursey wrote: >I just posted a first draft of the Quiescence MPI interface proposal. >The wiki page is here: > http://svn.mpi-forum.org/trac/mpi-forum-web/wiki/Quiescence > >Let me know what you think. Sorry it took so long to post. > >Cheers, >Josh > >_______________________________________________ >mpi3-ft mailing list >mpi3-ft(a)lists.mpi-forum.org >http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-ft

1 0