Search Mailing List Archives


Limit search to: Subject & Body Subject Author
Sort by: Reverse Sort
Limit to: All This Week Last Week This Month Last Month
Select Date Range     through    

[openflow-spec] [openflow-discuss] Proposal for a slicing mechanism for OpenFlow

Justin Pettit jpettit at cs.stanford.edu
Wed Aug 19 10:31:31 PDT 2009


> There is a difference between queueing and port-grouping. Ports are
> individual entities and can be configured as a group at anytime.
> Grouping of queues takes place at the configuration of the queue, i.e.
> we setup the same queue at a subset of ports, and then we can refer to
> these queues with the same identifier. It's not  a grouping though,
> since each of them is referred with a unique {port_id,queue_id} tuple.
> Of course, this could be implemented by a more generic grouping layer,
> but I guess it adds more complexity..

I understand how port-grouping works.  The reason to use 64-bit queue ids
hasn't been well justified, and I was throwing out possible interpretations
of what you were saying, since I've seen different reasons for their use in
different messages.

So based on this description, why do you want 64-bit queue ids?  It sounds
like they're going be tied directly to a port.  (Man, I hope I don't get
labeled as that guy who said, "Four billion queues out to be enough for
anyone.")

--Justin






More information about the openflow-spec mailing list