NIC Bonding speed vs LAG
 

Hi,
I've installed Wheezy on a new webserver for my office (I would have used stable but my RAID was not recognised)
I want to maximise the network throughput by bonding the two NICs (Intel Corporation 82576 Gigabit).
I've done this on other servers (using "Stable") with good results.
I tested the network speed using iperf.
With a normal setup (MTU 9000) I got;
0.0-10.0 sec 1.16 GBytes 990 Mbits/sec

Setup with Bond Mode 0 (MTU 9000) I get;
0.0-10.1 sec 323 MBytes 267 Mbits/sec

Not the result I'd had hoped for!

Both NICs go to a Netgear GS748TS stack.
I tried setting up a LAG on the stack (Static and LACP) but the network would not come up.

Anyone seeing anything like this?

TIA,
Steve

Sounds like your switch isn't doing the LACP bonding correctly and is getting confused seeing the same IP address mapped to NICs on different ports.

This might be helpful if you've not come across it already - http://backdrift.org/howtonetworkbonding

Hi TenTenths,

Thanks. I hadn't see that page before and it was helpful.
I managed to get my Ubuntu 12.04 LTS server running Mode 4 (server)/LACP (switch)
Is there a way to get figures for the maximum throughput of the server/switch? I've been using iperf which gives variable results, presumably because of the dynamic nature of Mode 4.

However, on a Debian Stable server I can't get it to run in the same configuration - only Mode 0 (server)/Static LAG (switch)

I don't know if this is because the version of the Ethernet Channel Bonding Driver;
Ubuntu 12.04 LTS - v3.7.1 (April 27, 2011)
Debian Stable - v3.5.0 (November 4, 2008)

The two servers are different, so it may be down to the builtin ethernet card chipsets. Is there a way to determine which modes are supported by the NIC card?

Your help is very much appreciated.

Steve

I honestly haven't a clue, my experience of bonding NICs is based around doing active/passive pairing rather than bandwidth boosting, I just happened to remember that switches needed to be compatible for active/active pairing.