On 9/5/07, Sean Hefty <[EMAIL PROTECTED]> wrote:
> >It's not a requirement to wait for the QoS determination of the SA
> >before issuing the SA PR requests.
>
> It sounds like you're suggesting that the end nodes just automatically gather
> and print the capabilities of the SA.

Already is a diag tool (saquery) which does this.

> (Why limit it to QoS only then?)

Only because it was the topic of discussion and the most "interesting"
capability to query.

> An administrator could run a separate tool to gather this information, rather 
> than
> it being done on all nodes automatically.  If the QoS determination is tied 
> into
> the behavior of the PR queries, then I think you end up either queuing the
> requests or responses somewhere.

Sure if they are somehow tied together.

> >Failover is pretty straightforward. The same query is made when SM LID 
> >changes.
>
> Yes - but existing paths that are in use may now suddenly provide or no longer
> provide QoS.

Indeed and maintaining QoS across SM failover is a hard problem which
is left to the individual SM implementations (as are other similar
failover issues).

-- Hal

> >Will the QoS code handle the new error status code which suggests a
> >different QoSClass or is it currently being handled like other errors
> >? Guess that could be a separate patch.
>
> This would be a separate patch, and is left up to each ULP at the moment.  For
> ipoib, I used the TClass, so the new QoS fields are not an issue.  For the
> rdma_cm, the QoS info is set by the user when using IPv4 addressing, so they
> would need to take appropriate action.
>
> - Sean
>
_______________________________________________
general mailing list
[email protected]
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general

To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general

Reply via email to