[Click] Queue Capacity and RatedUnqueue
Giovanni Giardinelli
giovanni.giardinelli at tin.it
Mon Jun 21 11:56:32 EDT 2004
Hi Eddie, you wrote on 19/06/2004 6.18.50:
The reason for which I use the RatedUnque element is that, in a DiffServ
environment, I want to control the EF traffic, and at the same time I want to ensure
that this kind of traffic will be served at the RATE parameter fixed in the configuration
of the RatedUnqueue element.
Moreover, also if I send at a rate much more low than 200,000 packets per second,
I still observe this behavior.
Could be for the limits of my router in processing 64 bytes packets?
Giovanni
>Giovanni,
>
>Do you still get drops with a simple "Unqueue" (not RatedUnqueue)?
>
>The rate on that RatedUnqueue is pretty slow: 200,000 packets per second. It's
>not surprising that your traffic generators can generate more load than that.
>
>Eddie
>
>
>
>Giovanni Giardinelli wrote:
>> Hi Eddie,
>>
>> I use three ethernet Gbit
>>
>> In order to generate packets, I use the element FastUDPSource.
>>
>> tester's configuration is the following:
>>
>> src2::FastUDPSource (204800,8192000,64,00:30:48:28:ba:28,192.168.10.145,
>> 1234,00:30:48:28:ba:ae,192.168.10.209,1234)
>> -> SetIPDSCP(46)
>> -> ToDevice(eth2);
>>
>> src3::FastUDPSource (20000,800000,64,00:30:48:28:ba:29,192.168.10.225,
>> 1234,00:30:48:28:ba:af,192.168.10.209,1234)
>> -> SetIPDSCP(0)
>> -> ToDevice(eth3);
>>
>> FromDevice(eth0)
>> -> c0 :: Classifier(12/0806 20/0001,-);
>>
>> c0[0]
>> -> ar0 :: ARPResponder(192.168.10.209 00:07:e9:1f:b2:50)
>> -> SimpleQueue
>> -> ToDevice(eth0);
>>
>> c0[1]
>> -> cf0:: Counter()
>> -> Discard();
>>
>>
>> The principal part of the configuration of the Router is the following:
>>
>> c0 :: Classifier(12/0806 20/0001,
>> 12/0806 20/0002,
>> 12/0800,
>> -);
>>
>> c2 :: Classifier(12/0806 20/0001,
>> 12/0806 20/0002,
>> 12/0800,
>> -);
>>
>> c3 :: Classifier(12/0806 20/0001,
>> 12/0806 20/0002,
>> 12/0800,
>> -);
>>
>> src0 :: PollDevice(eth0)
>> -> [0]c0;
>> src2 :: PollDevice(eth2)
>> -> [0]c2;
>> src3 :: PollDevice(eth3)
>> -> [0]c3;
>>
>> .............
>> .............
>> .............
>>
>> c2[2]
>> -> Paint(3)
>> -> Strip(14)
>> -> CheckIPHeader(1.255.255.255)
>> -> ipc2 :: IPClassifier(ip dscp 46, ip dscp 0);
>>
>> c3[2]
>> -> Paint(4)
>> -> Strip(14)
>> -> CheckIPHeader(1.255.255.255)
>> -> ipc3 :: IPClassifier(ip dscp 46, ip dscp 0);
>>
>> qef :: SimpleQueue(32000);
>>
>> ipc2[0]
>> -> qef;
>>
>> qef
>> -> RatedUnqueue(204800)
>> -> rt;
>>
>> ...........
>> ...........
>> ...........
>>
>> PokeHandlers( wait 5,
>> read qef.length,
>> read qef.droppes
>> loop
>> )
>>
>>
>> The qef.lenght=32000
>>
>> the qef.droppes increases in continuation
>>
>>
>>
>> Thanks
>>
>> Giovanni
>>
>>
>>
>> ----- Original Message -----
>> From: "Eddie Kohler" <kohler at CS.UCLA.EDU>
>> To: "Giovanni Giardinelli" <giovanni.giardinelli at tin.it>
>> Cc: <click at amsterdam.lcs.mit.edu>
>> Sent: Friday, June 18, 2004 11:04 PM
>> Subject: Re: [Click] Queue Capacity and RatedUnqueue
>>
>>
>>
>>>Giovanni
>>>
>>>Please send a more complete configuration. What are you using to
>>>generate packets?
>>>
>>>
>>>Eddie
>>>
>>>On Jun 18, 2004, at 2:06 PM, Giovanni Giardinelli wrote:
>>>
>>>
>>>>Hi all,
>>>>I'm doing some tests for a simple router architecture which have to
>>>>perform differentiated service between two kind of flows.
>>>>When I use a Queue element and a RatedUnqueue element in this way:
>>>>SimpleQueue(32000) -> RatedUnqueue(X), where X is a certain rate
>>>>value, even if the tester emission rate is <=X, the queue
>>>>always overflow, causing a lot of packets to be dropped. I use 64
>>>>bytes packets.
>>>>I have two question:
>>>>1) Why the click in my machine not allow me to fix a CAPACITY
>>>>parameter value for the queue greater than 32000 packets?
>>>>2) Why the RatedUnqueue element seems to not work correctly?
>>>>
>>>>_______________________________________________
>>>>click mailing list
>>>>click at amsterdam.lcs.mit.edu
>>>>https://amsterdam.lcs.mit.edu/mailman/listinfo/click
>>>
>
More information about the click
mailing list