Measuring RTP QoS params from SIPp load test

asked 2019-03-06 08:37:01 +0000

user32167 gravatar image


I was running a very small SIPp load test: sipp -sf uac.xml -i [xxx] -p [yyy] [uas-ip]:[yyy] -t t1 -m 500 -l 100 -r 5

with packet capture running in the background: tcpdump -i andy -s0 -vvv -w uac.pcap

Then after the load test is done, I turned off the tcpdump and ran it through tshark: tshark -nr uac.pcap -q -z rtp,streams

I notice that the QoS measures are pretty out of wrack: - Packet loss is in the negatives - Max delta is 21 ms - Max Jitter is 20K ms - Mean Jitter is 3K ms

The instances I ran on are very powerful EC2 instances. It makes me wonder if my setup's performance is really shitty or there's something wrong with the way I measure QoS on a SIPp load test

edit retag flag offensive close merge delete


Do you have RTP packets arriving out of order?

Anders gravatar imageAnders ( 2019-03-06 09:04:08 +0000 )edit

I am not sure - how do you check?

user32167 gravatar imageuser32167 ( 2019-03-07 06:48:39 +0000 )edit

RTP Seq values should increase by one in each packet. I think out-of-order and duplicated packets may confuse the analysis code.

Anders gravatar imageAnders ( 2019-03-07 09:47:39 +0000 )edit