>> Also >> the current level of the bonding driver precludes real balancing of outgoing >> data over the nics.
>This part I don't agree with. One trap I almost fell into was just looking at >the raw numbers of bytes transmitted for >each interface. Then I realized >that a fair amount of traffic flows over one of the interfaces before the >bonding >happens, because of the way I was testing. So, I took note of the >numbers, ran a test (14MB or so), and took the delta >of before and after. >When I did that, the numbers were almost identical. This was true of both the >round robin and xor >modes. Well, that depends on your setup. Please see my results posted some items back. You do agree transferring several TB does constitute a more thorough test than just 14MB. Also about 100 different nodes were involved in that test. In my case the switch are two physical Cisco switches configured as one logical switch (don't ask me the details, I am no Cisco guy, that is what they tell me). To summarize, the results for outgoing were: eth0 46606.8 Mb eth1 209.0 Mb. I call that unbalanced. Best regards, Pieter Harder pieter.har...@brabantwater.nl tel +31-73-6837133 / +31-6-47272537 Brabant Water N.V. Postbus 1068 5200 BC 's-Hertogenbosch http://www.brabantwater.nl Handelsregister: 16005077 ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
