Can only the 100Mbs part of a 10/100Mbs router fail?

Benjamin Scott bscott at ntisys.com
Tue Dec 21 22:14:01 EST 2004


  I realize this thread is old, but I'm catching up and I might be able to
offer some insight here...

On Wed, 8 Dec 2004, at 8:32am, travis at scootz.net wrote:
> Most consumer level 10/100 "switches" or "routers" are switching hubs.  
> That is there's a 100mb hub and a 10mb hub and the switching happens
> between the 10 and 100mb parts, not between each port. So that part might
> also be broken.

  I assume you're using "hub" to mean "repeater" like most do.  (The term
"hub" properly means "the center of any star topology network", which is
damn confusing since *BASE-T /is/ a star topology.)  What you're describing
as a "switching hub" is really the same as any ordinary repeater, just with
a learning bridge instead of a dumb bridge.

  You see, in reality, all Ethernet repeaters are single-speed.  That's
inherent in the nature of a repeater: Electrical signal comes in, is
re-timed, and repeated on all the other ports.  There's no buffering to
handle speed changes.  10BASE-T and 100BASE-T are two different signals,
after all.

  A "dual-speed repeater" is really two repeaters: A 10 megabit repeater and
a 100 megabit repeater.  A dumb bridge connects the two, so that traffic
that comes in on the 100 megabit repeater will be bridged and transmitted on
the 10 megabit repeater.  Physically, this is all one unit, of course (if
not one IC chip), but conceptually, that's how it works.

  It is very possible for the bridge in a dual-speed repeater to fry while
leaving the individual repeaters working.  One particular model of 8-port
repeater from LinkSys seemed to have a tendency to fail in this way.  It
once manifested itself to us as all the 100 megabit workstations in an
office working fine, but the 10 megabit network-attached printer was
unreachable.  Damn, but that took forever to figure out.

  The difference between a dumb bridge and a learning bridge: A dumb bridge
just forwards every data frame it gets.  Frame comes in on one interface, it
goes out all the others.  These are rather rare, and are used mainly for
"media conversion".  A learning bridge tries to keep track of what layer two
addresses (MAC addresses) it sees on which interface.  If it gets a frame
for an address it knows, it only sends it out the interface it saw that
address on.

  Back in the days when Ethernet ran on coax instead of twist-pair, learning
bridges were used to keep busy parts of the network well-isolated at layer
two.  These days, of course, we have switches: Multi-port learning bridges.

  I can also say that many of the consumer switches being sold at Staples,
Best Buy, and the like act like "real" switches, in terms of how they
forward frames.  Traffic on one port does not come out the other ports once
the destination address is learned.  This I've seen with a sniffer.  I have
no doubt that some "switches" work this way, but "most" might be a bit of a
stretch.

-- 
Ben Scott <bscott at ntisys.com>
| The opinions expressed in this message are those of the author and do  |
| not represent the views or policy of any other person or organization. |
| All information is provided without warranty of any kind.              |




More information about the gnhlug-discuss mailing list