AW: AW: namd-ibverbs fails to start

From: Norman Geist (norman.geist_at_uni-greifswald.de)
Date: Wed Nov 23 2011 - 00:16:28 CST

As far as he knows what kind of network he actually has and if rdma is
required.

;)

Norman Geist.

-----Ursprüngliche Nachricht-----
Von: Axel Kohlmeyer [mailto:akohlmey_at_gmail.com]
Gesendet: Dienstag, 22. November 2011 15:02
An: Norman Geist
Cc: David Hemmendinger; Namd Mailing List
Betreff: Re: AW: namd-l: namd-ibverbs fails to start

On Tue, Nov 22, 2011 at 1:49 AM, Norman Geist
<norman.geist_at_uni-greifswald.de> wrote:
> NOW I see! 10Gbit/s-Ethernet is NOT Infiniband! What you need to run RDMA
> over Ethernet is OpenMX (mMyrinetExpress). You can get from the net. But
> it's not easy to get it working.

actually, the _real_ bottom line is, that one cannot use
the precompiled NAMD binaries with that setup, but has
to compile an MPI version based on the existing OpenMPI
installation. no need to mess up the system with some
other protocol layer.

axel.

> Good luck
>
> Norman Geist.
>
>
> -----Ursprüngliche Nachricht-----
> Von: David Hemmendinger [mailto:hemmendd_at_union.edu]
> Gesendet: Montag, 21. November 2011 17:04
> An: norman.geist_at_uni-greifswald.de
> Betreff: Re: AW: namd-l: namd-ibverbs fails to start
>
>        I think that I wasn't clear enough, not being very familiar
> with ibverbs infiniband.  We are using RoCE, which, as I understand it,
> runs the infiniband protocol over tcp.  The IBM tech who installed it
> on our cluster ran tests, which it passed, and I'm able to use it with
> OpenMPI 1.4.2.  My comment about the failure of ibv_rc_pingpong referred
> only to what happened before I knew that I needed to specify -g GID,
> a recently-added feature, according to Mellanox documentation.
>        So my question was whether this new feature would mean that
> I'd need to modify an ibverbs call in charmrun -- since we're running
> over 10GB ethernet, we don't have an ib0.
>        Thanks,
>          David
>
>
>>this doesn't seem to be a problem of namd or charmrun, rather than a
> problem
>>of your infiniband configuration/installation. If the tests shipped with
>>ofed fail, then there's something wrong. If you don't want to spend too
> much
>>time with the problem, use the ipoib driver to use the infiniband with ip
>>traffic. Then u can just use a udp (faster) or tcp (NET version, usually
>>slower than udp) version of namd over the ip over infiniband stack which
> for
>>me was faster than the native verbs. Another advantage is that you can use
>>every possible mpi application like that also. Keep in mind to change the
>>connection mode to connected, _not_ datagram and set the mtu to 65520.
>>
>>$> echo connected > /sys/class/net/ib0/mode
>>$> ifconfig ib0 mtu 65520
>>
>>If that doesn't work also, something is wrong with your ofed or
infiniband,
>>then we can look further
>
>
>

-- 
Dr. Axel Kohlmeyer
akohlmey_at_gmail.com  http://goo.gl/1wk0
College of Science and Technology
Temple University, Philadelphia PA, USA.

This archive was generated by hypermail 2.1.6 : Mon Dec 31 2012 - 23:21:00 CST