[openfabrics-ewg] Cisco SQA results so far for OFED 1.0 rc5

[Jeff Squyres (jsquyres)]

These seem like fairly non-IB-specific names to put in the $PATH.  For
example, "write_lat" could refer to anything (file I/O, sockets, etc.).

Could these executables be renamed to be something more indicative of
what they really are?  E.g., "ib_rdma_write_lat", "ib_rdma_read_lat",
...etc.?

I'm happy to file a bug about this if desired.


> -----Original Message-----
> From: openfabrics-ewg-bounces at openib.org 
> [mailto:openfabrics-ewg-bounces at openib.org] On Behalf Of 
> Tziporet Koren
> Sent: Wednesday, June 07, 2006 6:14 AM
> To: Scott Weitzenkamp (sweitzen); Pavel Shamis
> Cc: openfabrics-ewg at openib.org
> Subject: RE: [openfabrics-ewg] Cisco SQA results so far for 
> OFED 1.0 rc5
> 
> Note that rdma_lat is the oldest test that was written.
> We wrote a new set of performance tests that are maintained:
> 
> Latency tests:
> write_lat - RDMA write
> read_lat  - RDMA read  
> send_lat  - Send (ud & rc)
> 
> Bandwidth test:
> write_bw - RDMA write
> read_bw  - RDMA read
> send_bw  - Send (ud & rc)
> 
> 
> All these tests are installed under <prefix>/bin. You can run 
> them with
> -h to get the usage.
> 
> Tziporet
> 
> -----Original Message-----
> From: Scott Weitzenkamp (sweitzen) [mailto:sweitzen at cisco.com] 
> Sent: Tuesday, June 06, 2006 11:06 PM
> To: Tziporet Koren; Pavel Shamis
> Cc: openfabrics-ewg at openib.org
> Subject: RE: [openfabrics-ewg] Cisco SQA results so far for 
> OFED 1.0 rc5
> 
> Great!  We were also able to reproduce the problem using rdma_lat.
> 
> We've added rdma_lat to our smoke test, so we detect this kind of
> problem sooner.
> 
> Scott Weitzenkamp
> SQA and Release Manager
> Server Virtualization Business Unit
> Cisco Systems
>  
> 
> > -----Original Message-----
> > From: Tziporet Koren [mailto:tziporet at mellanox.co.il] 
> > Sent: Tuesday, June 06, 2006 10:13 AM
> > To: Tziporet Koren; Scott Weitzenkamp (sweitzen); Pavel Shamis
> > Cc: openfabrics-ewg at openib.org
> > Subject: RE: [openfabrics-ewg] Cisco SQA results so far for 
> > OFED 1.0 rc5
> > 
> > Hi All,
> > 
> > We found the problem - it was a change in compilations 
> flags that was
> > inserted while added support for PPC64.
> > We will fix it for RC6.
> > Thanks Scott for finding this issue.
> > 
> > Tziporet
> > 
> > -----Original Message-----
> > From: openfabrics-ewg-bounces at openib.org
> > [mailto:openfabrics-ewg-bounces at openib.org] On Behalf Of 
> > Tziporet Koren
> > Sent: Monday, June 05, 2006 7:44 PM
> > To: Scott Weitzenkamp (sweitzen); Pavel Shamis
> > Cc: openfabrics-ewg at openib.org
> > Subject: RE: [openfabrics-ewg] Cisco SQA results so far for 
> > OFED 1.0 rc5
> > 
> > We will run the verbs performance tests to make sure nothing 
> > happened to
> > the basic verbs performance.
> > I will publish results tomorrow
> > 
> > Tziporet
> > 
> > -----Original Message-----
> > From: openfabrics-ewg-bounces at openib.org
> > [mailto:openfabrics-ewg-bounces at openib.org] On Behalf Of Scott
> > Weitzenkamp (sweitzen)
> > Sent: Monday, June 05, 2006 6:33 PM
> > To: Pavel Shamis
> > Cc: openfabrics-ewg at openib.org
> > Subject: RE: [openfabrics-ewg] Cisco SQA results so far for 
> > OFED 1.0 rc5
> > 
> > > Hi,
> > > The default mtu size in mvapich-RC5 was changed to MTU1024 
> > > (it solves bw 
> > > issue on PCI_X). You can try  VIADEV_DEFAULT_MTU=MTU2048 
> > parameter on 
> > > your platforms (PCI_EX) and I believe that you will get the 
> > > rc4 latency 
> > > numbers.
> > 
> > I'll try that for MVAPICH, but that doesn't explain the 
> worse latency
> > for Open MPI and Intel MPI.
> > 
> > Scott
> > _______________________________________________
> > openfabrics-ewg mailing list
> > openfabrics-ewg at openib.org
> > http://openib.org/mailman/listinfo/openfabrics-ewg
> > _______________________________________________
> > openfabrics-ewg mailing list
> > openfabrics-ewg at openib.org
> > http://openib.org/mailman/listinfo/openfabrics-ewg
> > 
> 
> _______________________________________________
> openfabrics-ewg mailing list
> openfabrics-ewg at openib.org
> http://openib.org/mailman/listinfo/openfabrics-ewg
>