nightrecon at verizon.net
Wed Jul 15 22:39:26 UTC 2009
David Kelly wrote:
> Not directly FreeBSD related, but how much of a chance is there that two
> machines could communicate directly over 5,000 feet of cat5 with no
> special hardware?
> IIRC the classic ethernet problem limiting the distance between the
> farthest points on a network had to do with timing and collisions. If
> these two NICs are configured full duplex then it seems one would have
> no idea how far away the other was due to timing issues.
No. Ethernet uses a protocol design called Carrier Sense Multiple Access
with Collision Detect, or CSMA/CD. The maximum lengths are indeed related to
timing and the timing is a direct result of the propagation delay in the
medium. The velocity factor will be some percentage of the speed of light.
So the time it takes for the smallest Ethernet frame to get from the two
farthest nodes will determine a window in which the two most distant nodes
upon attempting a transmit can tell that a collision occurred and
retransmit. The node(s) attempting to recover from a collision condition
will each generate a random time back off in the hope that one will begin a
packet transmission not at the same time as the other.
The timing patterns of the frames are finite and not infinitely adjustable,
e.g. there are limits which will declare a packet was not received and a
resend is therefore required. What you will experience with 5,000 of Cat5 in
full duplex is these limits will always be exceeded and the endpoints will
believe no packets are arriving at their destinations and lock itself into a
continual resend loop. When both ends do this you will have essentially
either very little, or zero throughput.
The max distance for UTP is 328 ft. Divide the 5,000 by 328 and it will tell
you how many bridges, hubs, or switches you will need to regenerate the
signal. You may find devices purporting to 'range extenders', but even these
will have distance limitations requiring more than one. Foofaraw.
> 100baseT uses lower power drivers than 10baseT, so perhaps 10baseT would
> work better.
> In any case, have boxes of cat5 on order so as to find out myself.
Sounds like a waste of time. Single mode fiber can support GB speeds some as
far as 10km. Single mode fiber is what you want to look at for this
distance. I'm not as current with long haul wireless links, but you may also
find this could be done with the right wireless endpoints and good antennae,
albeit you won't get the speed single mode fiber is capable of.
More information about the freebsd-questions