Hi Chait,
Yes, I agree - per DCBX spec - that is correct. Thanks for
clarification.
However, in centralized management construct like OpenFlow (I hope my
interpretation is correct here), I am trying to understand whether people have
thought about "engineering" lossless paths using MIBs defined by IEEE (and not
using DCBX).
Thanks,
Manoj Wadekar
Chief Technologist,
Storage Solutions Group
QLogic Corporation
-----Original Message-----
From: [email protected] [mailto:[email protected]]
Sent: Friday, October 21, 2011 11:40 AM
To: [email protected]; Manoj Wadekar
Cc: [email protected]; [email protected]
Subject: RE: [openflow-discuss] Q: DCB support
Rob Sherwood wrote:
>
> On Fri, Oct 21, 2011 at 8:25 AM, Manoj Wadekar
> <[email protected]> wrote:
> > Two interesting follow-up points:
> > 1. DCB Protocols: In case of OpenFlow - does one need pt-pt protocol like
> DCBX to achieve "lossless" priority negotiated in the path or OpenFlow will
> programmatically achieve same results?
>
> While I'm not a DCBX expert, you could imagine using OpenFlow to steer
> traffic onto dedicated physical links or queues (virtual links) that
> ensured lossless priority.
Manoj's question asks if there is anything in OpenFlow that allows setting up
specific links to be lossless in the first place. I believe the answer to that
question is no. DCBX must still be employed at the link-level (over LLDP and so
on) to negotiate the "lossless" semantics of a link. OpenFlow allows specific
links to be selected (a la "circuit switching" though I'm a little leery of the
term) as the preferred pathway through the fabric for a "flow" (which can be
defined quite flexibly via OpenFlow constructs).
I'm interested in hearing if the above description is accurate...
Regards,
-Chait.
> > 2. FCoE functionality in switches/adapters: What extensions are supported by
> OpenFlow for this?
>
> I was saying that someone *could* implement this as an extension, but
> to the best of my knowledge, people have only talked about doing it
> and not actually acted on it.
>
> Hope this helps,
>
> - Rob
> .
> _______________________________________________
> openflow-discuss mailing list
> [email protected]
> https://mailman.stanford.edu/mailman/listinfo/openflow-discuss
_______________________________________________
openflow-discuss mailing list
[email protected]
https://mailman.stanford.edu/mailman/listinfo/openflow-discuss
