[openib-general] Performance Degradation with OFED v. Voltaire(lustre)

[Bernadat, Philippe]

So after a bit more testing, setting the route path mtu to 1024 before
the qp creation (rdma_create_qp()) seems sufficient.
 

Philippe

> -----Original Message-----
> From: Bernadat, Philippe 
> Sent: Monday, December 18, 2006 3:09 PM
> To: Bernadat, Philippe; Or Gerlitz
> Cc: Hal Rosenstock; Eitan Zahavi; Roland Dreier; 
> openib-general at openib.org
> Subject: RE: [openib-general] Performance Degradation with 
> OFED v. Voltaire(lustre)
> 
> Or,
> 
> I did manage to fix it my way, by inserting this same
> 
> route->path_rec->mtu = IB_MTU_1024;
> 
> Before/after qp creation
> Before/after accept
> Before/after connect
> 
> So not sure which one really fixes it.
> 
> Philippe
> 
>  
> 
> > -----Original Message-----
> > From: Bernadat, Philippe 
> > Sent: Monday, December 18, 2006 2:42 PM
> > To: Or Gerlitz
> > Cc: Hal Rosenstock; Eitan Zahavi; Roland Dreier; 
> > openib-general at openib.org
> > Subject: RE: [openib-general] Performance Degradation with 
> > OFED v. Voltaire(lustre)
> > 
> > I have tried both fiexs, none of these  improve performance ... 
> > Let me check the number of packets again.
> > 
> > Philippe 
> > 
> > > -----Original Message-----
> > > From: Or Gerlitz [mailto:ogerlitz at voltaire.com] 
> > > Sent: Monday, December 18, 2006 2:03 PM
> > > To: Bernadat, Philippe
> > > Cc: Hal Rosenstock; Eitan Zahavi; Roland Dreier; 
> > > openib-general at openib.org
> > > Subject: Re: [openib-general] Performance Degradation with 
> > > OFED v. Voltaire(lustre)
> > > 
> > > Bernadat, Philippe wrote:
> > > > 3) Is there a way to change the MTU from within the lustre 
> > > LND kernel
> > > > module. I saw that the IB perf programs did this with the 
> > > modify_qp()
> > > > APIs.
> > > 
> > > yes, go to the place where the lustre NLD active side gets 
> > > RDMA_CM_EVENT_ROUTE_RESOLVED event on its rdma cm id and then set
> > > 
> > > 	lustre_id->route->path_rec->mtu = IB_MTU_1024;
> > > 
> > > Or.
> > > 
> > > 
> > > 
> > >