[ofa-general] iWARP issues

Kanevsky, Arkady Arkady.Kanevsky at netapp.com
Fri Nov 2 07:39:11 PDT 2007


comments inline.

Arkady Kanevsky                       email: arkady at netapp.com
Network Appliance Inc.               phone: 781-768-5395
1601 Trapelo Rd. - Suite 16.        Fax: 781-895-1195
Waltham, MA 02451                   central phone: 781-768-5300
 

> -----Original Message-----
> From: Roland Dreier [mailto:rdreier at cisco.com] 
> Sent: Thursday, November 01, 2007 6:27 PM
> To: Kanevsky, Arkady
> Cc: bboas at systemfabricworks.com; OpenFabrics General
> Subject: Re: [ofa-general] iWARP issues
> 
>  > - iWARP Support for Peer-to-Peer Applications, this is CM  
> > interoperability issue
> 
> I guess the relevant people will be there from the RNIC 
> vendors, but this seems more like an IETF/rdma consortium issue to me.

yes, I expect RNIC vendors will be there.
The hope is that we can come up with changes in iwcm to ensure interop.
This is why is planery item.

It is true if IB vendors do not require to participate.

> 
>  > - iWARP + TCP host stack port space sharing (required by 
> IETF iSER  > spec.)
> 
> Maybe we can try to brainstorm for alternate solutions, but 
> it seems we are missing all the key stakeholders from the 
> Linux networking side to really resolve this.

That is the goal.

> 
>  > - missing verbs (IB-only, iWARP-only, and 
> iWARP-nonstandard),  >  for example FMR, send with 
> invalidate. (I recall that complete list was  > flash out  >  
> a year ago)  > (ULP changes to take advantage of these verbs: 
> e.g. NFS-RDMA, iSER).
> 
> I haven't seen any discussion of this yet.  Is it really 
> worth taking up face-to-face time on this?  It seems like it 
> should all be pretty straightforward, and even if it isn't, 
> we haven't spent the time to figure out what the not 
> straightforward parts are yet.

There was a discussion about it almost a year ago.
I think if RNIC vendors will fund a person to move this forward.
This is why it need planery time.
IB vendors are also needed there because many of the same verbs
are transport independent and defined in IBTA spec v1.2.

> 
>  > - RDMA connection timeout; expand RDMA_CM API to support 
> timeout  > paramater (not iWARP specific)
> 
> Again, I haven't seen any discussion yet, so it doesn't seem 
> worth taking up face-to-face-time until we know what the 
> sticking points are.

If Sean is doing it already we may not need planery time for it.

> 
>  - R.
> 



More information about the general mailing list