Re: Mcast packet loss 2.6.8.1 kernel

[Ahmed Kadwa]

breaking news :
output from iperf shows we are losing packets on udp as well for a
bandwidth setting above 4m.

./iperf -s -u -i 1
------------------------------------------------------------
Server listening on UDP port 5001
Receiving 1470 byte datagrams
UDP buffer size:  105 KByte (default)
------------------------------------------------------------
[  3] local 192.168.60.11 port 5001 connected with 192.168.60.13 port
1026
[ ID] Interval       Transfer     Bandwidth       Jitter   Lost/Total
Datagrams
[  3]  0.0- 1.0 sec  1.02 MBytes  8.54 Mbits/sec  0.361 ms    0/  726
(0%)
[  3]  1.0- 2.0 sec  1.28 MBytes  10.7 Mbits/sec  0.048 ms   67/  978
(6.9%)
[  3]  2.0- 3.0 sec  1.19 MBytes  10.0 Mbits/sec  0.043 ms    0/  850
(0%)
[  3]  3.0- 4.0 sec  1.19 MBytes  10.0 Mbits/sec  0.068 ms    0/  851
(0%)
[  3]  4.0- 5.0 sec  1.03 MBytes  8.63 Mbits/sec  0.037 ms  115/  849
(14%)
[  3]  5.0- 6.0 sec  1.19 MBytes  10.0 Mbits/sec  0.040 ms    0/  851
(0%)
[  3]  6.0- 7.0 sec  1.19 MBytes  10.0 Mbits/sec  0.060 ms    0/  850
(0%)
[  3]  7.0- 8.0 sec  1002 KBytes  8.21 Mbits/sec  0.244 ms    0/  698
(0%)
[  3]  8.0- 9.0 sec  1.19 MBytes  10.0 Mbits/sec  0.419 ms    0/  850
(0%)
[  3]  9.0-10.0 sec  1.19 MBytes  9.98 Mbits/sec  0.064 ms    0/  849
(0%)
[  3]  0.0-10.0 sec  11.5 MBytes  9.61 Mbits/sec  0.062 ms  182/ 8353
(2.2%)

./iperf -c p0e0 -u -b 10m
------------------------------------------------------------
Client connecting to p0e0, UDP port 5001
Sending 1470 byte datagrams
UDP buffer size:  105 KByte (default)
------------------------------------------------------------
[  5] local 192.168.60.13 port 1026 connected with 192.168.60.11 port
5001
[ ID] Interval       Transfer     Bandwidth
[  5]  0.0-10.0 sec  11.7 MBytes  9.81 Mbits/sec
[  5] Server Report:
[  5]  0.0-10.0 sec  11.4 MBytes  9.59 Mbits/sec  0.539 ms  187/ 8344
(2.2%)
[  5] Sent 8344 datagrams

Thus we see packet loss.
However when we set the client bandwith (-b 4m) all is fine, but
anything above this causes packet loss. I expect there to be very little
loss (<1%) under these conditions. The two nodes above are in the same
physical box, on separate processors, each running their own kernel
separated by a switch that only has traffic between the nodes.

output when tcpdump is run as well wtih no cmd line options

/flash/sbin # ./iperf -s -u -i 1
------------------------------------------------------------
Server listening on UDP port 5001
Receiving 1470 byte datagrams
UDP buffer size:  105 KByte (default)
------------------------------------------------------------
[  3] local 192.168.60.11 port 5001 connected with 192.168.60.13 port
1026
[ ID] Interval       Transfer     Bandwidth       Jitter   Lost/Total
Datagrams
[  3]  0.0- 1.0 sec   482 KBytes  3.95 Mbits/sec  0.321 ms    0/  336
(0%)
[  3]  1.0- 2.0 sec   485 KBytes  3.97 Mbits/sec  0.050 ms    7/  345
(2%)
[  3]  2.0- 3.0 sec   490 KBytes  4.01 Mbits/sec  0.054 ms    0/  341
(0%)
[  3]  3.0- 4.0 sec   488 KBytes  4.00 Mbits/sec  0.077 ms    0/  340
(0%)
[  3]  4.0- 5.0 sec   422 KBytes  3.46 Mbits/sec  0.154 ms    0/  294
(0%)
[  3]  5.0- 6.0 sec   488 KBytes  4.00 Mbits/sec  0.051 ms    0/  340
(0%)
[  3]  6.0- 7.0 sec   488 KBytes  4.00 Mbits/sec  0.122 ms    0/  340
(0%)
[  3]  7.0- 8.0 sec   425 KBytes  3.48 Mbits/sec  0.154 ms    0/  296
(0%)
[  3]  8.0- 9.0 sec   551 KBytes  4.52 Mbits/sec  0.304 ms    0/  384
(0%)
[  3]  9.0-10.0 sec   488 KBytes  4.00 Mbits/sec  0.058 ms    0/  340
(0%)
[  3]  0.0-10.0 sec  4.70 MBytes  3.94 Mbits/sec  0.059 ms    7/ 3357
(0.21%)

Packet loss is lower??


 
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html