OT:Switch boosts ethernet?

[Tony Wasson]

On Thu, Feb 19, 2009 at 11:21 AM, Alan Dayley <alandd at consultpros.com> wrote:
> If this were a casual situation, we might do what you suggest.  But
> using a cable that is too long is known to be a source of intermittent
> and odd problems.  In this situation we'd rather do it right once and
> not worry about errors happening when we least want them.
>
> This is a pretty good if not detailed explanation of the length limit:
> http://www.duxcw.com/faq/network/cablng.htm
>
> Basically, as I understand it, when the Ethernet sender ships out the
> packet it will wait for an acknowledge from the destination.  The
> signals running through the cable are very, very fast but are not
> instant to the other end.  When the cable is too long, the sender
> could declare a time out on the acknowledge and resend before the
> destination gets the packet.  Collisions or other problems could then
> ensue.  So, theoretically, it does not matter much how "good" the
> cable is because the limit is based propagation of the signals.

Also -- with long cables you get noise on the cable called cross talk
(NEXT & FEXT). The cross talk is worse on  a long cable at 100Mbps or
Gigabit speeds. If you are running half duplex, you will get late
collisions on a 'too long' cable that would cause noticeable packet
loss or packet delay that will show up as frame drops in a video feed.

A switch is a really good solution. Please make sure to use Full
Duplex for the smoothest packet flows. Fiber is a good solution if you
had to go a lot further, but there are higher costs with that.