[c-nsp] does duplex mismatch affect UDP throughput?
Scott Granados
scott at granados-llc.net
Sat Jul 23 23:02:48 EDT 2011
Duplex mismatches effect everything so far as I know. You're getting a lot
of late collisions and generally badness with frames which means retrans and
errors and really bad throughput. (many orders of slower throughput than
you'd expect)
D
-----Original Message-----
From: Martin T
Sent: Saturday, July 23, 2011 9:52 PM
To: cisco-nsp at puck.nether.net
Subject: [c-nsp] does duplex mismatch affect UDP throughput?
I have a following network topology:
T60[eth0] <-> [Fa0/2]WS-C2950C-24[Fa0/1] <-> [eth0]PE860
eth0 interface of T60 has autonegotiation enabled and comes up in
100BASE-TX/Full mode:
root at martin-ThinkPad-T60:~# mii-tool eth0
eth0: negotiated 100baseTx-FD, link ok
root at martin-ThinkPad-T60:~#
Fa0/2 interface of WS-C2950C-24 is configured to "speed auto"/"duplex
auto" and thus comes up in 100BASE-TX/Full as well:
Fa0/2 -> T60 connected 1 a-full a-100
10/100BaseTX
WS-C2950C-24 port Fa0/1 is configured to "speed 100" and "duplex full":
Fa0/1 -> PE860 connected 1 full 100
10/100BaseTX
..and as PE860 eth0 interface is configured to autonegotiate it's
speed, there is a duplex mismatch between the switch and PE860
machine.
I configured IP address 10.10.10.1 to eth0 of T60 and 10.10.10.2 to
eth0 of PE860. There is no packet loss between the T60 and PE860 if
there is no traffic other than ICMP "echo requests" and "echo replies"
between the hosts:
root at martin-ThinkPad-T60:~# ping -qc1000 -i 0.1 10.10.10.2
PING 10.10.10.2 (10.10.10.2) 56(84) bytes of data.
--- 10.10.10.2 ping statistics ---
1000 packets transmitted, 1000 received, 0% packet loss, time 100039ms
rtt min/avg/max/mdev = 0.190/0.595/1.207/0.253 ms
root at martin-ThinkPad-T60:~#
Now if I set up an Iperf server in TCP mode in PE860 machine(iperf -s
-fm) and execute Iperf client in T60:
iperf -c 10.10.10.2 -fm -d -t20
..I get rather poor results:
[ 5] local 10.10.10.1 port 51381 connected with 10.10.10.2 port 5001
[ 4] local 10.10.10.1 port 5001 connected with 10.10.10.2 port 46702
[ ID] Interval Transfer Bandwidth
[ 5] 0.0-20.0 sec 190 MBytes 79.4 Mbits/sec
[ 4] 0.0-20.1 sec 2.84 MBytes 1.18 Mbits/sec
..and there are hundreds of CRC errors on switch Fa0/1 interface and
PE860 eth0 interface. Now if I execute Iperf server in UDP mode(iperf
-s -fm -u) and Iperf client with following settings:
iperf -c 10.10.10.2 -fm -d -t20 -u -b 100m
[ 4] local 10.10.10.1 port 54559 connected with 10.10.10.2 port 5001
[ 3] local 10.10.10.1 port 5001 connected with 10.10.10.2 port 34610
[ ID] Interval Transfer Bandwidth
[ 4] 0.0-20.0 sec 229 MBytes 95.9 Mbits/sec
[ 4] Sent 163099 datagrams
[ 3] 0.0-12.1 sec 1.33 MBytes 0.93 Mbits/sec 40.014 ms 925/ 1876 (49%)
[ 3] 0.0-12.1 sec 1 datagrams received out-of-order
[ 4] Server Report:
[ 4] 0.0-20.0 sec 228 MBytes 95.5 Mbits/sec 0.009 ms 604/163098
(0.37%)
[ 4] 0.0-20.0 sec 1 datagrams received out-of-order
So I guess duplex mismatch does affect UDP traffic as well? And why
are upload/download so unequal?
regards,
martin
_______________________________________________
cisco-nsp mailing list cisco-nsp at puck.nether.net
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/
More information about the cisco-nsp
mailing list