[openfabrics-ewg] OFED 1.1 planning meeting - summary

[Tziporet Koren]

Hi all,

This is the outcome of the meeting we had today regarding OFED 1.1
schedule and features.

 

Tziporet

 

 

1. Schedule:

============

Target release date: 31-Aug 

Intermediate milestones:

1.    Create 1.1 branch of user level code and rc1: 27-Jul

2.    Feature freeze : 3-Aug 

3.    Code freeze (rc-x): 25-Aug 

4.    Final release: 31-Aug

 

In general all agreed but it seems aggressive schedule. We will delay in
1 week if needed or drop some features.

There was a request for another OFED release toward SC06 that will
include most updated Open MPI version and we agreed this is possible.

 

git tree of kernel code will be available on Sandia servers once their
system administrator will setup the server with git installed (should be
this week)

 

2. Features:

============

Note: features that are under low priority may not be qualified in final
release due to schedule limitations.

 

1. OS:

Novell:

- SLES 9.0 SP3*

- SLES10 (official release)*

Redhat:

- Redhat EL4 up3

- Redhat EL4 up4*

kernel.org:

- Kernel 2.6.17*

 

* Changes from last release

 

Note: Redhat EL4 up2, Fedora C4 and SuSE Pro 10 were dropped from the
list.

We will keep the backport patches for these OSes and make sure OFED
compile and loaded properly but will not do full QA cycle.

 

2. General changes:

-        lib32 on 64 bits systems

-        Kernel code based on 2.6.18

-        HCA fatal - full flow support - Low priority

-        High Availability in IPoIB and SRP

-        Bug fixes

 

3. OSM (new code based on the trunk):

-        Partition Manager (Pkey) - Low priority 

-        Pre-computed routing load from file 

-        Primitive QoS - As technology preview 

Mainly developed and verified by Voltaire.

 

4. SDP:

-        Beta quality (higher stability)

-        Improved latency

-        Improved bandwidth of small messages (Naggle algorithm) - done

-        Support the backlog parameter in the listen call

-        support sending/receiving out of band data 

-        Interoperability with previous SDP implementation

We need SDP from Cisco to test the interoperability with their SDP

 

5. SRP:

-        GA quality

-        DM (Device Mapper) - for high availability

-        Basic failover/failback testing with daemon+srp+XVM/MPP and
Engenio target

A technical mail was published on the general list. (Subject: Needed
changes to support fail-over drivers). Need help from Roland to close
the technical details since he is SRP maintainer.

 

6. IPoIB:

-        High availability support using a daemon in user level

 

7. uDAPL:

-        Scalability features needed for Intel MPI 

-        Going to take the new code from the trunk

 

8. OSU - MVAPICH:

-        Based on 0.97 (+ bug fixes)

-        Message coalescing

 

9. Open MPI:

-        Open MPI 1.1.1 - Depending on the dates/schedule of OFED 1.1
and Open MPI 1.1.1 (If not then Open MPI 1.1 will be used)

-        The major differences between Open MPI 1.1 and 1.1.1 can be
seen here: http://svn.open-mpi.org/svn/ompi/trunk/NEWS

 

10. MPI tests:

-        Replace to the new test versions from LLNL, Intel, OSU

 

11. iSER:

-        Stability - code review and bug fixes at iser and libiscsi code
related to error handling (libiscsi is a service module used by both
iscsi_tcp & iser)

-        Testing more platforms (e.g. ppc64 and ia64)

-        Performance improvements

-        The libiscsi fixes are (2.6.18-rc2/3) and will (2.6.19) be
pushed upstream and from there be propagated to distros (eg SLES10  RH5)
through their merge process.

 

12. RDS:

-        Oracle and SilverStorm should update 

Need to decide if RDS should be removed from OFED since Oracle does not
support it for now.

Sujal will check it and we will get to decision soon.

 

13: Management: 

-        Madeye utility 

-        Add saquery tool

-        Enhancement to ibnetdiscover tool with grouping function

-        New ibutils package: 

o        Port error counter check 

o        Port performance counters dump 

o        Link width and Link Speed check by flag

 

 

 

 

 

 

-----Original Message-----
From: openfabrics-ewg-bounces at openib.org
[mailto:openfabrics-ewg-bounces at openib.org] On Behalf Of Sujal Das
Sent: Monday, July 17, 2006 8:32 PM
To: openfabrics-ewg at openib.org
Subject: [openfabrics-ewg] OFED 1.1 planning meeting

 

Hello all,

 

We would like to call an OFED v1.1 planning meeting ASAP.  We can use

the regular time and call-in number that Shawn/Jeff from Cisco had set

up earlier.  Shawn/Jeff: will you please confirm and send a reminder?

 

Some critical feature requirements for OFED 1.1 are based on customer

feedback on OFED 1.0, specifically some missing features that continue

the perception of OFED 1.0 as being inferior to proprietary stacks:

 

- Support for SLES 9 in general

- SRP failover and MPP

- SDP usability

- RDS usability

- UDAPL scalability

- High Availability

- Additional OS/CPU platforms support for iSER, sync with upstream

kernel

- Etc

 

Thanks,

 

Sujal  

 

_______________________________________________

openfabrics-ewg mailing list

openfabrics-ewg at openib.org

http://openib.org/mailman/listinfo/openfabrics-ewg

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openfabrics.org/pipermail/ewg/attachments/20060725/423b182f/attachment.html>