On Mon, Jun 14, 2004 at 06:17:51PM +0100, Martin Michlmayr wrote:
> I just tested on my RaQ2 and I get up to 2500kB/s via FTP. This is
> in my LAN with a 100 MBit switch.
Hmm. Maybe this is really the maximum...
> Which network device do you use (the first or second)? What's your
> setup? Do you get such low rates also when you simply do FTP without
> any masquerading or other stuff?
Okay. This is the current state:
I could finally compile the current CVS kernel + the Cobalt
patches from http://www.colonel-panic.org/cobalt-mips/
The Raq2 is now running 2.6.7-rc3. With this upgrade
everything got bettter, but it's still not perfect.
With the standard Debian ftpd connections begin with about
2.5MB/s, but then drop to 1.37MB/s. Using vsftpd it's
similar, but the transfer rate doesn't drop so much.
I can get about 1.67MB/s. There is no real difference
whether firewalling/masquerading is used at the same time
or not. I'm just a bit confused because I was told by
another Raq2 owner that he could get transfer rates of about
6 or 7MB/s using the current CVS kernel. I did all my
tests by transferring the ungzipped gcc-3.4.0 tarball.
That's 182MB. The nice thing is that I can now
transfer a big file via ftp without killing the whole
machine. With 2.4.26 beginning an ftp transfer made
all other connections unuseable. Outgoing SSH
connections dropped down so much that I could wait
for every typed character to appear. Maybe it's
a good idea to file a Debian bug report. Their
standard kernel seems to be problematic.
Anyway, the network problems didn't go away completely.
I had one "almost" network freeze with the new kernel
resulting in lots of dropped packages and transfer rates
similar to what I described in my first mail. Restarting
the network solved this.
tessa:~# ifconfig eth0
eth0 Link encap:Ethernet HWaddr 00:10:E0:00:31:F5
inet addr:192.168.2.10 Bcast:192.168.2.255 Mask:255.255.255.0
inet6 addr: fe80::210:e0ff:fe00:31f5/64 Scope:Link
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:4314429 errors:2 dropped:10412 overruns:0 frame:5
TX packets:3013429 errors:23 dropped:0 overruns:4 carrier:19
RX bytes:1151870219 (1.0 GiB) TX bytes:1305479776 (1.2 GiB)
My current setup is like this: I have connected eth0 to
a 100Mbit switch and eth1 to a gateway router via 10MBit.
The box has about 1 day uptime now and works stable at
the moment. I'm waiting for the next network freeze...
ANother thing worth noting is that after upgrading to 2.6.7-rc3
I have the same problem with "df" that was described on this
Filesystem 1K-blocks Used Available Use% Mounted on
df: `/': Invalid argument
df: `/proc': Invalid argument
df: `/sys': Invalid argument
df: `/dev/pts': Invalid argument
df: `/dev/shm': Invalid argument
df: `/boot': Invalid argument
df: `/export': Invalid argument
strace gives the same info as it was described in the past.
All in all, I'm very interested in any possible improvement.
I need a low cost server that is silent enough that I can
sleep in the same room and this Cobalt seems like a very
good candidate (after replacing the small fan of course).
thanks for your help.
I'm not on the list. I'm reading this list via a web-gateway.
de-moc-ra-cy (di mok' ra see) n.
Three wolves and a sheep voting on what's for dinner.