The IP over ATM Mailing List Archive by date

[Juha Heinanen <Juha.Heinanen@lohi.dat.tele.fi>]

   Let's get real, as Curtis pointed out.  ATM networks have a single point
   of failure now called the switch.  If you buy into ATM networks now, you
   buy into the single point of failure model.

i have tried to stay away from this, but the above is alittle too much.
it is perfectly normal to have computers of one LIS connected to several
switches, like is the case in my own LIS.  if one of those switches (or
one port of one switch) fails, the whole LIS should not need to fail,
but unfortunately that is currently the case if the only ARP server of
the LIS happens to be connected to the failed switch or port.

-- juha