VanLUG Email Archive
Mike Bell (athasian@altavista.net)
Sun, 15 Aug 1999 15:11:27 -0700
On Sat, Aug 14, 1999 at 01:59:25PM -0700, Ya`akov N Miles wrote:
> I have heard about "switches" and I have a "hub" for my 100 mbits/second
> EtherFast LAN. Can someone explain the difference between a "switch"
> and a "hub", also, if two computers are communicating via he hub, does
> that make all the rest of the ports on the hub unusable until the two
> communicating computers finish exchanging packets?
OK, a hub emulates the behavior of 10base2, ie signals sent out from machine
A disseminate to all the other ports. As a result, each machine has to
listen before sending in case someone else is sending a packet. However,
since the signals travel over the cable at a finite but really fast speed,
the latency in the LAN cannot be too great, otherwise you get two collision
zones, where both might send a packet and assume it made it safely, without
ever realizing they collided. This results in packets having to be
retransmitted when TCP catches on, very slow.
Anyway, a switch sends the signal only to the host for which the signal was
intended, it has limited intelligence with which to determine this, hence
costs more than a hub. So with a switch, you are no longer sharing 100mbit
with the other machines on the network, each machine has 100mbit each way
all to itself.
As for your last question, you might have gleaned it from the above, but
basically each machine waits until the current packet is done, then tries to
sneak it's packet in before another is sent. If two machines send at once,
both wait a random interval then try again. Sort of inefficient, but it
works.
-- Mike Bell
This archive was generated by hypermail 2.0b3 on Thu 09 Sep 1999 - 06:40:34