Horst, collisions *don't* happen on switchs. (in general. They do, but generally
don't).
Hubs are subject to collisions, which is *why* we use switches.
One of my servers plugged into one of my main switches:
uptime
16:31:10 up 21 days, 23:40, 1 user, load average: 1.00, 1.00, 1.00
ifconfig
eth0 RX packets:8427002 errors:0 dropped:0 overruns:4274 frame:0
TX packets:5605652 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:100
RX bytes:2816996680 (2686.4 Mb) TX bytes:741618198 (707.2 Mb)
My laptop which has been up for almost a month, plugged in to both a cable modem (on a
hub-type network) and one of my main switches at work:
ifconfig
eth0 RX packets:7719491 errors:0 dropped:0 overruns:0 frame:0
TX packets:5400248 errors:70 dropped:0 overruns:1492 carrier:0
collisions:374 txqueuelen:100
RX bytes:1284777968 (1225.2 Mb) TX bytes:3076203002 (2933.6 Mb)
Hmm, I think those totals are a little low, they may have been reset, but you see low
or no collisions. Some of the errors, overruns or collisions on my laptop may be due
mostly to the cable modem.
Hubs are subject to collissions because all the ports are wired together. A simple
hub is only a nexxus of network cables, meaning 1 conversation at once. (ie a party
line). A switch allows for ~x conversations at once where X is the number of ports.
Now this isn't completely true for all switches. For instance a 4-cross bar switch
only allows for 4 conversations even on a 24-port switch. Two of cross bars may be
taken up by full duplex, per conversation (ie two systems talking back and forth). So
on a 4-cross bar switch, you may only have two full duplex conversations per unit of
switch time (ie how long it takes to switch the cross bar to another port.)
I think newer switches are moving away from crossbarring and using something like
queing and forwarding (although I forget the exact term for it).
Horst, if your nic is plugged into a switch with that many collisions, try a new nic,
a new port and/or a new cable. Reset the collisions count and watch it. You may have
a short.
On Tue, Aug 14, 2001 at 02:52:21PM -0700, Horst Lueck wrote:
> Cory, thanx for comments and tips, now:
> ...
> > eth0: DC21143
> > eth0 Link encap:Ethernet HWaddr 00:A0:CC:D3:C1:E7
> > collisions:295789 txqueuelen:100
> >
> > !!! You have this nic plugged in to a switch!! 295,789 collisions which
> > should be 0!
> >
> You lost me here on the switch (hub?) issue. (one more sentence please)
> There are 3 machines on that local switch with frequent traffic so I
> assumed some collisions are normal. Those locals are the eth1 - not the
> eth0.
> But the eth0 are on 'www' (since they didn't cause the hangup I haven't
> looked at what is down-stream).
>
> Since this maybe just a misunderstanding I didn't see a need to increase
> the Eug-LUG bandwidth this reply ............... Horst.
>
>
