tcp troughput weirdness

[Alexis Yushin]

Hmm,

Recently I've also been seeing less than what I'd expect tcp throughput
on FreeBSD 5.4R machines. I've got six 5.4R boxes with dual Gigabit em
interfaces. netperf gives me:

Recv   Send    Send                          
Socket Socket  Message  Elapsed              
Size   Size    Size     Time     Throughput  
bytes  bytes   bytes    secs.    10^6bits/sec  

 65536 131072 131072    10.01     551.10   


However scp/ftp etc doen't go above 200Mbit. I'd expect higher numbers?

Alexis

ps I've tried all of the sysctl and mtu tweakings... No change. Interesting
is that netperf  lo0 gives me about 2.5Gbit, but scp performs at about 128Mbit
while system CPU load is only 20%. What shall I tweak further?

On Tue, Jul 12, 2005 at 02:07:38PM +0300, Danny Braniss wrote:
> > > combining
> > > 	sysctl net.inet.tcp.sendspace=131072
> > > and
> > > 	sysctl net.inet.tcp.inflight.enable=0
> > >
> > > did the trick!
> > 
> > Congratulations!  But I wonder why the throughput of FreeBSD=>Linux
> > was almost equal to that of Linux=>FreeBSD.  If the settings above
> > improves the throughput of FreeBSD=>FreeBSD, the throughput of
> > FreeBSD=>Linux would also be improved with them.  Is it improved?
> > 
> i did the sysctl on the client/sender :-)
> 
> > > now can someone remind me what inflight does?
> > 
> > In my understanding, it tries to estimate bandwidth-delay product and
> > tries to avoid injecting too much data segments into networks.
> > So, if it underestimates bandwidth-delay product, throughtput may be
> > reduced.
> > 
> > Regards,
> > Noritoshi Demizu
> 
> 
> _______________________________________________
> freebsd-stable at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-stable
> To unsubscribe, send any mail to "freebsd-stable-unsubscribe at freebsd.org"