Minutes for FTB conference call - 2009 Jul 8th

Attendees

• Argonne National Lab: Rinku
• Oak Ridge National Lab: Aniruddha, Thomas, David, Hoony, Josh Hursey (from IU)
• Ohio State University: Karthik, Mani
• Indiana University: Torsten, Abhishek
• University of Tennessee: Absent
• Lawrence Berkeley National Lab: Absent

Items Discussed

1. FTB Update
   1. Rinku went over the presentation on the FTB 0.6.1 release. RC is scheduled to release on July 9th and final release is tentatively on July 15th. We discussed features and plans for FTB-0.7 as well.
   2. Rinku will send the CIFTS BOF text (for SC'09) for everyone's review.
   3. We discussed event aggregation. It was decided that event throttling will be performed for all FTB-enabled components. For RAS-based FTB-enabled components on Cray and BG/P, both ANL and ORNL are investigating methods for event aggregation and root cause analysis. David mentioned that we can extend this to Linux machines, if some software (like Ganglia) can provide us sufficient information/logs to determine rules and patterns.
2. ORNL
   1. Ground work for the fusion application is completed. ORNL folks will start working on the FTB-enabled part of the Fusion application. A publication is expected in a month or so.
   2. Thomas will be working with Hoony on Cray. They will be tested FTB 0.6.1 on the Cray machines. Hoony will provide Rinku the details of the bug encountered with FTB on Cray.
3. OSU
   1. Karthik informed that MVAPICH 1.4 is now FTB-enabled and available in the MVAPICH trunk for testing. Release expected in 2 weeks.
   2. OSU will test FTB 0.6.1 with the FTB-IB 1.0 software and update the FTB-IB webpage accordingly.
4. IU
   1. Abhishek informed that they are continuing work on the Open MPI notifer component (which performs some event aggregation), which will eventually integrate in FTB.