Minutes for FTB conference call - 2009 Sep 30th
From CIFTS
Attendees
- Argonne National Lab: Absent
- Oak Ridge National Lab: Hoony, Scott, Thomas, Aniruddha
- Ohio State University: D.K., Sonia
- Indiana University: Josh, Abhishek
- University of Tennessee: Absent
- Lawrence Berkeley National Lab: Absent
Items Discussed
- FTB update:
- FTB 0.6.1 was released on Sep. 14th 2009. All major bugs have been fixed. Rinku and Hoony are aware of some minor issues that would be addressed for the next minor release.
- ICPP Update:
- Rinku gave a presentation about CIFTS, which drew a good attention. Some questions received included (FTB self-resiliency, monitoring faults in the system using FTB)
- Pete introduced CIFTS project during his keynote talk at P2S2 workshop. The title of his talk was "Challenges for system software fon exascale platforms"
- Indiana State Update
- Abhishek tested FTB release and found it running well.
- Abhishek and Josh brought up "FTB_Connect" exit issue which Abhishek had posted on mailing list. Abishek and Josh pointed out that FTB should not arbitrarily terminate users' application. It should instead notify user application of specific error it encounters. Hoony would discuss the issue with Rinku, and take this into consideration for the next release.
- The Indiana University group has put together a website dedicated to CIFTS activities: http://osl.iu.edu/research/ft/cifts/ . This web site contains information on FTB events thrown/caught by OpenMPI, how to enable FTB support in Open MPI, and how to test out the installation. The first round of FTB support is currently available in the Open MPI development trunk, and scheduled for inclusion in the v1.5 release series. This site will be updated as IU develops/refines the events and protocols/workflows supported by Open MPI.
- Josh asked to link the web site from CIFTS website.
- Plan for SC demo needs to be finalized
- ORNL Update:
- Anirudda reported his work on multitasking capability is maturing, and will start with FTB shortly. Anirudda has a plan for SC demo, but no details yet.
- Scott reported his work on LAMMB/OpenMPI/FTB. He is in the stage of fixing bugs. No particular plan for SC demo.
- Thomas spent some time looking for possible useful cases for pub/sub capability of FTB. His automatic testing suite for FTB is nearly completed (using OpenMPI MTT).
- Ohio State Update:
- Ohio State Team tested the final RC version of FTB 0.6.1, which ran fairly well. They plan to test the regular FTB 0.6.1 release shortly.
- Sonia, a new postdoc of D.K. Panda, has joined the CIFTS Team.
- MVAPICH2 is now FTB-enabled. Details can be found at the website.
- New/Updated Contents for SC BOF are expected from every participating institute. Details will be announced soon.