Version 2.4.3.  I just tested 2.4.4, and I have the same problem.  

> 11 Mb/s is ok for 100 Mbit network, if you account for:
> 	- network overhead
> 	- kernel overhead
> 	- disk overhead

When I switch back to kernel 2.2, I can get around 80Mbit/sec to hosts on
the same switch.  I'm going to have to install too many boxes if I can only
get 11Mbit/sec out of them.  I have 5 boxes running kernel 2.2 that serve
out around 100 million images a day combined.  I'm not using 2.4 in
production anymore because of random box crashes, and network performance
issues.  But, I'd like to be able to use it, especially for when Apache 2.0
is released.  And all of the latest distros are shipping with it also.  





> -----Original Message-----
> From: Florin Iucha [mailto:florin at iucha.net]
> Sent: Thursday, May 03, 2001 1:48 PM
> To: 'tclug-list at mn-linux.org'
> Subject: Re: [TCLUG] kernel 2.4 network performance probs
> 
> 
> On Wed, 2 May 2001, Austad, Jay wrote:
> 
> > I have tried 3 different ethernet cards, a tulip, an 
> eepro100, and a 3c59x.
> > I cannot get more than about 11Mb/sec out of them.  At 11Mb/sec,
> 
> 11 Mb/s is ok for 100 Mbit network, if you account for:
> 	- network overhead
> 	- kernel overhead
> 	- disk overhead
> 
> > /var/log/syslog fills with this:
> > May  2 17:51:57 i5 kernel: Undo loss 10.10.220.101/3419 c2 
> l0 ss2/2 p0
> > May  2 17:52:00 i5 kernel: Undo loss 10.10.220.101/2634 c2 
> l0 ss2/2 p0
> > May  2 17:52:00 i5 kernel: Undo loss 10.10.220.101/3228 c2 
> l0 ss2/2 p0
> > May  2 17:52:06 i5 kernel: Undo loss 10.10.220.101/4646 c2 
> l0 ss2/2 p0
> > May  2 17:52:06 i5 kernel: Undo loss 10.10.220.101/4647 c2 
> l0 ss2/2 p0
> > May  2 17:52:09 i5 kernel: Undo loss 10.10.220.101/4222 c2 
> l0 ss2/2 p0
> > May  2 17:52:09 i5 kernel: Undo loss 10.10.220.101/4524 c2 
> l0 ss2/2 p0
> > May  2 17:52:14 i5 kernel: Undo loss 10.10.220.101/4520 c2 
> l0 ss2/2 p0
> 
> What version of 2.4?
> 
> > 10.10.220.101 is the machine I'm testing from.  The message 
> appears from any
> > machine I'm testing on.  All machines are plugged into the 
> same switch, and
> > forced to 100/full.  I have no problems with my 2.2 
> kernels, only the 2.4
> > kernels.  Does anyone have any ideas?
> >
> 
> florin
> 
> _______________________________________________
> tclug-list mailing list
> tclug-list at mn-linux.org
> https://mailman.mn-linux.org/mailman/listinfo/tclug-list
>