[Click] IPClassifier

Eddie Kohler kohler at cs.ucla.edu
Mon May 3 15:30:33 EDT 2004


Udayan Deshpande wrote:
> Hi Eddie,
> 
> Thanks for your reply...
> 
> The delay is hovering around 500ms (for resumption of packet
> forwarding). I'm using Click 
> 1.3pre1.
> 
> Udayan Deshpande

500ms!?!?  That is very surprising, other people on this list have
seen much slower delays (they are able to hotswap several hundred
times a second).

I'm no longer supporting Click 1.3pre1.  Please try the current
anonymous CVS sources and let us know if the delay is still that
high.  Hotswaps should be faster now.

Eddie


> 
> Eddie Kohler wrote:
> 
> 
>>Hi Udayan,
>>
>>It's too bad that hotswap introduces these delays.  How much delay
>>are you experiencing?  And on what version of Click?  Hotswap should
>>be faster on the current anonymous CVS version of Click than it was
>>previously.
>>
>>To answer your question:  There is currently no way to change Classifier
>>or IPClassifier's behavior other than hotswap.
>>
>>Eddie
>>
>>
>>On Apr 23, 2004, at 12:56 PM, Udayan Deshpande wrote:
>>
>>
>>>Other than hotconfig is there any way of of changing the behaviour of
>>>IPClassifier at
>>>runtime? For example,
>>>
>>>making a the classifier behave as
>>> IPClassifier(dst tcp port 8000, dst tcp port > 1023, tcp);
>>>from
>>> IPClassifier(dst tcp port 7000, dst tcp port > 1023, tcp);
>>>
>>>so that the  "dst tcp port > 1023" flow is uninterrupted while the
>>>first one is modified.
>>>
>>>The hotconfig introduces delays that are unacceptable to my
>>>application. Any suggestions
>>>would be appreciated.
>>>
>>>Udayan Deshpande
>>>_______________________________________________
>>>click mailing list
>>>click at amsterdam.lcs.mit.edu
>>>https://amsterdam.lcs.mit.edu/mailman/listinfo/click
>>



More information about the click mailing list