Hi,
before sending out new question, let me answer my old question (for the
archives):
Carsten Aulbert wrote:
We did some data transfer tests. In the simplest setup we connected two
nodes back to back (direct cable connect) and created a TCP stream in
each direction via netperf for 60s (default parameters, TCP_STREAM).
[...]
After some discussions with (co-) authors of netperf and e1000, we
reached the conclusion that this test is not valid. Either one should
use the TCP_STREAM on one end and TCP_MAERTS on the other or TCP_RR with
multiple "bursts", i.e. compile netprf with --enable-burst and run with
the test with -- -b 4 for example.
With this we were able to get very good performance measures throughout
all mtu settings, albeit we needed MTU of 3000+ to get "line" speed.
Sorry for the late answer.
Carsten
--
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
