[Mono-list] Why is MONO network streaming slow?

[Will Murnane]

On 6/19/07, Adam Tauno Williams <adam at morrison-ind.com> wrote:
> > wow there is so much wrong with this test,
> > First I find it hard to believe that you got up to 90% efficiency on a
>
> I noticed that too.
> "NET --> .NET : 94 Mbit/s   O.K." on a 100Mb/s network.
> Bogus.  No way.
On a 100mbit switched network with half a dozen computers going
through a single 100mbit link to a 48-port switch through another
100mbit link to another half dozen, here's a test from one machine to
another:
[me at firsthost]$ ./ttcp -t -l1400 -n1048576 -s otherhost
ttcp-t: nbuf=1048576, buflen=1400, port=2000
ttcp-t: socket
ttcp-t: connect
ttcp-t: 0.2user 2.1sys 2:05real 1% 0i+0d 0maxrss 0+1pf 15315+23csw
ttcp-t: 1468006400 bytes processed
ttcp-t:   2.41663 CPU sec  =    593222 KB/cpu sec,  4.74578e+06 Kbits/cpu sec
ttcp-t:   125.738 real sec =   11401.5 KB/real sec,   91212.1 Kbits/sec

91 is pretty close to 94.  And other things are happening on this
network, too.  My home box pushes 11.9 MB/s = 95.2 megabits via iscsi,
because it's an uncrowded network and a simple protocol.  Sure, you
don't get that via (say) SMB, but this test's not doing anything
complicated.

> >  transfer  at all, network over head should have knocked down your
> > transfer more than that.
What overhead?  Please be more specific.