On Sunday 01 May 2005 14:19, Luca Landi wrote: > Alex Hermann ha scritto: > > >> server: eth0 192.168.1.1 <---> client1 : 192.168.1.101 > > >> eth1 192.168.1.2 <---> client2 : 192.168.1.102 > > >> eth2 192.168.1.3 <---> client3 : 192.168.1.103 > > >> eth3 public ip <---> inet > > Why do you need such a garbled topology? you want your server to act as a > switch (a Layer-2 switch) but you configured it much like as if it was a > router (a Layer-3 switch), that is, with an IP address assigned to each > of its interfaces. To my humble opinion there is a misconception here: if > you want your server to play the role of the switch then all you have to > do is to bring up a bridge interface, which means to enslave the physical > devices (eth0,1,2) to a logical one (traditionally a logical net device > named something like "br0") and then assign ONE single IP address to the > *logical* device only.
Sadly enough, some things aren't possible with the (linux) bridging code. My connections are a horrible mix of Gbit, 100Mbit, 10Mbit and wifi. To get the max out of the Gbit connection I need a high MTU on that iface. The bridging code doesn't support a MTU other than 1500 at all (or just not across different bandwidth devices). Also, there seem to be some more problems with my nics and bridging that need some further investigation. I still think that my setup is most suitable for my situation and allows for maximum performance on every iface. It seems this setup is not common enough to warrant implementation in dnsmasq, so I will look for some workaround. Alex.
pgp12jfnnowm3.pgp
Description: PGP signature
