Try using netperf (http://www.netperf.org/) too. I've found it to be an
extremely valuable tool.
Charles
-----Original Message-----
From: Niek Bergboer [mailto:[EMAIL PROTECTED]]
Sent: Thursday, April 05, 2001 4:35 PM
To: [EMAIL PROTECTED]
Subject: Network throughput tuning
Hi,
I run two systems on an intranet. The intranet itself is rather large,
but the two machines in question are connected to the same 100 Mbps/FDX
switch. I would like to optimize network throughput for Machine 1.
Machine 1 is a AMD K6-2 233 w/ 64 MB RAM running FreeBSD 4.2-STABLE from
around mid march and has a dc NIC. Machine 2 is a dual Celeron 466 running
Linux 2.4.2, and also has a dc NIC ("de4x5" driver in Linux terms).
In order to measure network throughput, I make sure _not_ to use the
disk I/O subsystem and issue the following commands:
machine1:~$ rsh machine2 dd if=/dev/zero bs=1048576 count=128 > /dev/null
(Linux doesn't understand bs=1m)
which yields between 9.0 and 9.2 MB/s which looks good.
machine2:~$ rsh machine1 dd if=/dev/zero bs=1m count=128 > /dev/null
gets me between 7.6 and 7.8 MB/s while this used to be 8.4 MB/s when
machine1 was still running Linux.
In short: the BSD machine receives 9.1 MB/s and sends 7.7 MB/s. Not that
I'm complaining, and the lower send rate may well be due to the Linux
box not handling the incoming stream well, but my question is: Did I
do _everything_ on the BSD box to ensure maximum throughput?
The tuning I did is:
sysctl -w kern.ipc.maxsockbuf=2097152
sysctl -w net.inet.tcp.rfc1323=1
sysctl -w net.inet.tcp.sendspace=1048576
sysctl -w net.inet.tcp.recvspace=1048576
Thanks in advance,
Niek
--
Conscience doth make cowards of us all.
-- Shakespeare
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-hackers" in the body of the message
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-hackers" in the body of the message
