Re: Ethernet Channel Bonding

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]


WipeOut wrote:
Jim Christiansen wrote:

Hello Everyone,

I've just joined the list and need to bite the bullet and tryout Fedora. I've got about 100 RedHat9 boxes being served NIS and NFS'ed home from one RH9 box. The one main server also serves 30 thin-clients.

I have posted this to the K12LTSP list as well, so I know that a few of you may recognize this mesage- sorry for repeats.

I want to increase the network capacity on the main server by adding another nic to the same subnet. I'm not a networking or Linux pro, but this short article may do the trick for me.

The article speaks about one nic recieving and one nic transmitting- sounds OK. Then it continues by stating that each nic plugs into its own hub or switch- OK too, if the switches can be linked.

I ask for your comments and really appreciate any help.



Jim, I doubt it would use one card to recieve and another to send.. In my mind that would be pointless.. it would mean you were simulating a full duplex ethernet connection with 2 half duplex conenctions..

What is more logical is that it would load balance over the two cards or route return traffic back over the card that recieved it... either way both cards should send and recieve to gain a benefit of increased throughput.. but anyway give it a try and see how it goes..

Here is another alternative source specifically the sections on split access and load balancing..

Jim's referring to "bonding", where you make 2 100Mb NICs look like a single 200Mb NIC (or give you two 100Mb NICs in a failover mode). The trick is that the switch (or host) that the NICs connect to MUST support bonding. If they don't, then don't bother with this and go the gigabit mode.

Before you do ANYTHING, you should first make sure you really are
saturating your NICs (that's hard to do with 100Mb NICs).  Use something
like mrtg or the like to see just what the NICs are doing on your main
server.  I'll bet you find you're compute- or disk-bound on that server.
- Rick Stevens, Senior Systems Engineer     rstevens@xxxxxxxxxxxxxxx -
- VitalStream, Inc.              -
-                                                                    -
-               The Theory of Rapitivity: E=MC Hammer                -
-                                  -- Glenn Marcus (via -

[Index of Archives]     [Current Fedora Users]     [Fedora Desktop]     [Fedora SELinux]     [Yosemite News]     [Yosemite Photos]     [KDE Users]     [Fedora Tools]     [Fedora Docs]

  Powered by Linux