Natasha Gude wrote:
> David Erickson wrote:
>> Dan Wendlandt wrote:
>>  
>>> David Erickson wrote:
>>>    
>>>> Hmm, I'm not sure that bindings_storage will work here since this is 
>>>> a time sensitive lookup, holding back packets from a flow waiting 
>>>> for a callback could be very detrimental, what do you think?
>>>>
>>>> -David
>>>>         
>>> Hi David and Glen,
>>>
>>> We had a quick chat here about how to best meet your needs and 
>>> decided that the best approach is actually the simplest:  you can 
>>> just have your own component listen to all packet in events to fill a 
>>> python dictionary of IP-to-location mappings based on the packet-in 
>>> event's data.
>>> Granted, this won't be super scalable in a large network, but if you 
>>> need real scalability, you will probably want to port the app to C++, 
>>> in which case you can easily access the same info via Authenticator.  
>>> Does that sound reasonable?
>>>     
>>
>> Dan,
>> Thanks for the reply. We currently do have code that creates and 
>> tracks ip:mac:dpid mappings, the problem was getting this information 
>> back in to authenticator for when it creates the flow_in events. 
>> However I think what we can do is listen for flow_in's ourself, and if 
>> we determine the destination chosen by authenticator is not optimal 
>> (ie it selected the gateway when we can actually go directly to an 
>> ip), then we will issue a STOP to the event, and post our own flow_in 
>> event with the fixed up destination connector object.
>>
>> That will handle getting routing to set the correct route, then as a 
>> quick hack we can add a member variable to the Flow_in_event object to 
>> contain a dest mac for rewriting, which can be examined by the 
>> handle_flow_in within Routing at the last hop in the Full Route, and 
>> add the action to rewrite.
>>
>> Does that sound reasonable?
>>
>> Thanks,
>> -David
>>
>> _______________________________________________
>> nox-dev mailing list
>> [email protected]
>> http://noxrepo.org/mailman/listinfo/nox-dev_noxrepo.org
>>   
> Instead of re-posting, you could also const cast the event and modify 
> dst to point to a different connector inline.  Other than that, it 
> sounds reasonable to me.
> 
> Natasha
> 

Ah good point.

Thanks,
David

_______________________________________________
nox-dev mailing list
[email protected]
http://noxrepo.org/mailman/listinfo/nox-dev_noxrepo.org

Reply via email to