nc/dd vs iperf

[Neil Watson]

Greetings,

I'm performing some network tests on some virtual machines (KVM). I've
performed tests using Iperf and dd/nc. I would expect the results to be similar
yet they are not. Tests are performed between two identical RHEL5, 64BIT
virtual hosts. The hosting hosts are also RHEL5, technically a hypervisor, but
are both identical. Both hosts/HV's are connected by a 1Gb connection on same
switch stack (HP Procurves).

Iperf
Server:
/usr/local/bin/iperf -s

Client:
iperf -c 10.125.70.2 -t 300

VM to VM on separate HV's
[ 3] 0.0-300.0 sec 26.1 GBytes 746 Mbits/sec
[ 3] 0.0-300.0 sec 26.1 GBytes 747 Mbits/sec

dd/nc
Server:
nc -vvnl 6000 > /dev/null

Client
dd if=/dev/zero bs=1k count=26M|nc -vvn 10.125.70.2 6000

VM to VM on separate HV's
27917287424 bytes (28 GB) copied, 1349.3 seconds, 20.7 MB/s (165Mb/s)
27917287424 bytes (28 GB) copied, 1257.03 seconds, 22.2 MB/s (178Mb/s)

Shouldn't these two tests show similar results? 

-- 
Neil Watson
Linux/UNIX Consultant
http://watson-wilson.ca
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists