On Thu, Oct 13, 2011 at 04:23:58PM -0700, Dan Talayco wrote:
> Recently a question came up about the number of responses that may result 
> from a request. I don't recall seeing anything in the spec about this.
> 
> There are places where this might be optional (each port stats or flow stats 
> entry could be put in a different message). But there may be places where 
> it's required. For example, when a flow stats request is made, there is a 
> possibility that the size of the response is too big for one OF message (the 
> header length field is 16 bits and there are switches that support tens of 
> thousands of flows). I'd be interested in others' experiences and thoughts. A 
> couple of notes:
> 
> * Correct me if I'm wrong, but nothing prevents the switch from always 
> responding with small OF messages, e.g., one entry per OF message. Should 
> this be advised against?
> 
> * All the messages in response to the request will have the same transaction 
> ID. Controllers need to be prepared for this.
> 
> => Is there an issue here in that we can't identify the last message 
> associated with the transaction? A barrier request could be used by the 
> controller; presumably the barrier reply would get queued behind the previous 
> transaction response.  Might be considered expensive for every stats query.

A non-stats request has 0 or 1 reply messages.  A barrier can be used
to verify that the request had 0 replies.

A stats message has 1 or more reply message.  Every message before the
last one has the "more" flag set.  The last message does not have the
"more" flags set.
_______________________________________________
openflow-discuss mailing list
[email protected]
https://mailman.stanford.edu/mailman/listinfo/openflow-discuss

Reply via email to