[Gllug] Networking: Would this work?

[John Hearns]

On Tue, 2002-05-28 at 22:39, Bruce Richardson wrote:
> On Tue, May 28, 2002 at 08:06:38PM +0100, Murray wrote:
> > But I *was* wondering what the limit to this would be.  It doesn't matter to
> > me, as I only have a tiddly small network, but I assume at some point, you
> > are going to notice your network speeds drop if you have too many hubs all
> > daisychained together.  But I'm not sure quite where that limit would be?
> 
> It isn't the number of hubs, it's the amount of traffic on the segment
> (and the number of hosts creating that traffic).  As that rises, so do
> the number of collisions till a point is reached where devices spend
> more time backing off and retransmitting than they do transmitting data.
> This puts a practical upper limit on the capacity of an ethernet network
> that is significantly below its theoretical maximum.
> 
> There's also a maximum length to an ethernet segment.  Maybe some wise
> soul can give a url to some stats, I'm going from memory.

In fact, the OReilly chapter is on-line:
http://www.oreilly.com/catalog/enettdg/chapter/ch13.html


> The above product boasts "address
> table up to 1K MAC addresses" - connecting 1K machines into a switched
> network would be pushing it somewhat in most circumstances. Switches
> are better than hubs in this respect of course.

What Ian says is right, but maybe not clear.
AFAIK it is OK to put lots of machines on a switched network - because
not all of the 1000 MAC addresses are learned by all switches.
Remember that all a switch is is a multi-port bridge.

If it's OK with everyone, maybe I'll do a quick write-up about
repeaters, hubs and bridges later on today.






-- 
Gllug mailing list  -  Gllug at linux.co.uk
http://list.ftech.net/mailman/listinfo/gllug