[Asterisk-Users] re hardware requirement - asterisk

[Rich Adamson]

> > fxp0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500
> >          address: 00:02:55:30:54:28
> >          media: Ethernet autoselect (100baseTX full-duplex)
> >          status: active
> >          inet 192.168.1.1 netmask 0xffffff00 broadcast 192.168.1.255
> >          inet6 fe80::202:55ff:fe30:5428%fxp0 prefixlen 64 scopeid 0x1
> > xl0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500
> >          address: 00:01:02:78:11:e8
> >          media: Ethernet autoselect (10baseT)
> >          status: active
> >          inet 203.219.167.126 netmask 0xfffffffc broadcast
> > 203.219.167.127
> >          inet6 fe80::201:2ff:fe78:11e8%xl0 prefixlen 64 scopeid 0x2
> > 
> > For fxp0, the internal interface although the nic can do full-duplex
> > it 
> > seems to me that it is only running simplex!!
> 
> Why do you think it is running simplex.  I read the above and see
> where it says "(100baseTX full-duplex)"
> 
> I don't think 10BaseT can run full duplex.  I could be wrong but
> I don't think so.
> 
> But why does it matter?  A single VOIP connection will not even
> use 1% of a simplex 10BaseT.  Simplex 100BaseT should be able
> to handle dozens and dozens of calls

Just for fun, I moved our * box to a 10meg al-cheapo hub to force 10-half,
placed a sip-to-sip call (via two C7960's) and noticed audio was very
much half duplex. Very irritating to say the least (worse then most digital
cell-to-cell calls).

Then without changing anything other then moving the * interface to an
upstream switch running 100 full (and verifying settings), the 
half-duplex-sounding audio effects completely disappeared (as expected).

While both tests were being conducted, I ran a Sniffer analyzer to monitor
packets and validate results.

10-half vs 10-full does have a substantial impact on quality. Moving from
10-full to 100-full would have no impact unless I could have loaded it
with more rtp sessions then what I currently have the ability to do.
And, FWIW, an interface set to half-duplex on one end with full-duplex
on the other end was by far worse then when both ends of the cat 5 matched.

All tests were conducted by forcing rtp traffic "thru" * (didn't allow
the rtp to flow between the two sip phones).

BTW, 10-Full setting is truly available on a large number of NICs, but
not all. Obviously, the older stuff didn't support it, nor do the older
Cisco 10 meg interfaces, etc.