[ewg] EWG/OFED meeting minutes fo June 29 09

[Tziporet Koren]

These are the meeting minutes for EWG/OFED meeting for June 29 09

Meeting summary
=============
1. OFED 1.5 pre-alpha release will be done next Monday (July 6)
2. OFED 1.4.2: Will be done with critical Lustre and NFS/RDMA fixes
only. Will have only partial QA.
3. RDMAoE: We will wait  to see the progress with Linux kernel

Details:
======
1. OFED 1.5 feature freeze date
Decided to wait with feature freeze till we know what is the status of
RDMAoE and new ipath driver.
All other components are ready for feature freeze (major features are
in, bug fixes will be done later)
In order to enable testing to start we will create pre-alpha release on
Monday next week.
AIs: 
- Qlogic to provide a date for ipath driver to be ready
- Mellanox to work with kernel maintainers on RDMAoE
- Tziporet to check the status of management package with Sasha the
maintainer

2. OFED 1.4.2 - there is a request from Sun to have bug fixes only 
release since there are critical bugs that prevent Lustre to run over
OFED 1.4 and 1.4.1
Decision: We will do 1.4.2 but only with the needed bug fixes. Only
partial QA will run not to impact OFED 1.5 release.
Jon for OGC will look into the bugs.
Note: The interop was already done with 1.4.1 and will not be done for
1.4.2. 
         1.4.2 will be recommended only for customers that will need the
new bug fixes only.

3. RDMAoE:
There was a long discussion on RDMAoE in OFED 1.5
I will not summary all the discussion, since many items were already
raised on the list.
Outcome: We wish to see progress with the kernel maintainers. 
               Decision will be delayed in few weeks based on this
progress.
We do not wish to start voting on features 

4. MPI in OFED: 
Decision in XWG was not to change for 1.5 and bring it as a BOF in SC09
with the customers
MPI versions to be in OFED 1.5:
- Open MPI: 1.3.x or 1.4 (based on Open MPI latest version)
- MVAPICH: 1.2
- MVAPICH2: 1.4

New MMU notification: 
- New code is being tested with Roland. There is still a bug with
Fortran that is under debug.
- Kernel code was not yet submitted to the kernel. 
- After issue is resolved and if new code will be submitted we can
consider if to take it to OFED too

Tziporet