[Live-devel] Strange output of openRTSP -Q

[Norbert Donath]

Hi,

I had a strange behaviour running openRTSP -Q. num_packets_lost showed
a very high number:

begin_QOS_statistics
server_availability 100
stream_availability 100
subsession video/H263-2000
num_packets_received 41043
num_packets_expected 20158
num_packets_lost 4294946411
elapsed_measurement_time 600.006624
kBytes_received_total 13964.541000
measurement_sampling_interval_ms 1000
kbits_per_second_min 0.000000
kbits_per_second_ave 186.191824
kbits_per_second_max 447.190117
packet_loss_percentage_min -2480.000000
packet_loss_percentage_ave 0.000000
packet_loss_percentage_max 2.000000
inter_packet_gap_ms_min 0.033000
inter_packet_gap_ms_ave 14.215230
inter_packet_gap_ms_max 4374.570000
subsession audio/AMR
num_packets_received 5487
num_packets_expected 2804
num_packets_lost 4294964613
elapsed_measurement_time 600.006624
kBytes_received_total 1476.055000
measurement_sampling_interval_ms 1000
kbits_per_second_min 0.000000
kbits_per_second_ave 19.680516
kbits_per_second_max 103.613901
packet_loss_percentage_min -3100.000000
packet_loss_percentage_ave 0.000000
packet_loss_percentage_max 60.000000
inter_packet_gap_ms_min 0.089000
inter_packet_gap_ms_ave 106.221868
inter_packet_gap_ms_max 5920.050000
end_QOS_statistics

As you can see I added debug output for num_packets_expected. For some
reason there where more packets received than expected. I assume that
some packets where received more than once. I also have the tcpdump
output but I didn't see double sequence numbers at a quick look (it's
16 MB and too big to analyse in detail).

I had a look in the code and I think openRTSP counts all received
packets regardless if they are valid or not. So it would be an
explanation for the numbers.

The high number for num_lost_packets is therefore a negative number as
the difference is printed as unsigned. I would suggest to convert the
difference to int to be sure to have a reasonable output. I modified
it for testing in a local version and it printed a negative packet
loss. I think providing a patch is not worth the effort ;-)


Regards,
Norbert Donath


-- 
Visit us at "3 GSM World Congress 2005" in Cannes, France
February 14th - 17th, stand F36, hall 2

SIGOS Systemintegration GmbH
 - TESTING IS OUR COMPETENCE - 
Fon +49 911 95168-0
www.sigos.de