" ... The configuration screen doesn't have anything to indicated this is supported and I don't recall seeing anything in the manual about it. So it's probably not supported ..."
It is supported in kernel 2.4.18 that comes with RH 7.3.
In /etc/ sysconfig/network-scripts you create a file called ifcfg-bond0
You also make changes to ifcfg-eth0 and ifcfg-eth1, where the NICs
are 'slaved' / bonded to bond-0.
When booting, bond-0 is brought up first, fallowed by the NICs. From
then on they share MAC and IP addresses. So it looks like just one NIC.
You can have as many bond's/NICs as you have open slots for. As
one bond with several NICs, or several bonds with multi NICs.
" ... Are you doing this fornd network reduancy? ..."
Yes.
And no.
Yes, it does supply reduancy. But it also allows you to have/utilize more
bandwidth. And no, I use it mainly for the bandwidth. It dosen't "add"
bandwidth, it just allows for better use of the bandwidth that is there.
One of the NICs is usually/always ready, so when the switch sends a packit
there is less latency.
Where I use it, is on the Linux Cluster I have/am building. And this is where I want to
use NASLite. As a NAS appliance for the cluster. The cluster will be used in a video
wall, multi screens, each showing a part of the whole graphig etc. Or as a killer
game server for doing network gamming. Or both. Whatever I use it for I want/need
to keep down the latency of the network. Even Gig-a-bit networking can make gains
this way.
I do know that it is a kernel function. But I don't know if it is just
a RH thing or all/every distro can use/do it. And I'm not sure if the kernel
source code is included with NASLite with the NASLite required additions.
So that if nessessery I could re-compile the kernel with the bonding included
in it myself, if it's not.
George
|