Hello Bruno

As the way sipp computes its statistics has been enhanced, may I ask you
if you still encounter the same problem with the 2.0 version ?

Regards,
Olivier Boulkroune
 


Message: 5
Date: Tue, 23 Jan 2007 21:22:47 +0100
From: Bruno Van Den Bossche <[EMAIL PROTECTED]>
Subject: [Sipp-users] Accuracy of time-measurements by SIPp
To: [email protected]
Cc: [EMAIL PROTECTED]
Message-ID: <[EMAIL PROTECTED]>
Content-Type: text/plain; charset=ISO-8859-15

Hallo all,

As we recently took up some SIP-performance testing again and since our
tests in the past the -trace_rtt option was added, we compared the
measurement results obtained using three different methods:

-trace_stat option which prints out global statistics about all calls.
-trace_rtt which prints out more detailed statistics about each
individual call
- a network capture using Wireshark

Our test-setup was the following:
- switched network
- dual core intel machines running at 2.8Ghz
- Read Hat Linux Enterprise Edition installed
- a call rate of 100 caps, 2 SIPp's directly communicating, UDP
- a slightly modified uac and uas scenario: actually the scenario's are
the default, but we increased the response-time repartition and
call-lenght repartition granularity to 1 ms
(<ResponseTimeRepartition value="1,2,3,4,5,6,7,8,9,10,11,12,...,300/>)

The measured results are as follows:
trace_stat:
        <1 : 19606
        <2 : 2559
        <3 : 1
        <4 : 1
        (average: 0.615)
                [ (19606*0.5 + 2559*1.5 + 2.5 + 3.5)/22167]

trace_rtt:
        <1 : 0
        <2 : 6948
        <3 : 15207
        <4 : 10
        <5 : 1
        <6 : 1
        average: 2.187

pcap:
        <1 : 22043
        <2 : 3
        <3 : 10
        <4 : 111
        average: 0.729

So we notice a very big difference between the output of -trace_stat and
-trace_rtt.  The output from the Wireshark trace and -trace_stat seems
to be more in line.  Especially considering there is already some
processing done by the time SIPp gets its hands on the packet.

Is this an expected result?  The difference between the -trace_rtt
output and trace_stat seems very large.  Although I initially expected
that the trace_rtt output would allow us to get more detailed diagnostic
results.

Another odd result is that although Wireshark reports 111 messages to
arrive later than 3ms, nor trace_rtt nor -trace_stat seem to detect
this.

Has anyone else noticed this behavior? And is it to be expected or do
you think it can be fixed?


Kind Regards,
Bruno

-- 
Bruno Van Den Bossche
[EMAIL PROTECTED]
http://www.ibcn.intec.UGent.be

*************

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
_______________________________________________
Sipp-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/sipp-users

Reply via email to