[f-nsp] Loadbalacing (Trunking/LACP) on foundry MLX: unequal load distribution.

[Gunther Stammwitz]

Hello colleagues,

I'm using a foundry mlx that connects to an IXP with 2 GIGEs.
We have created an LACP-Trunk but the traffic doesn't really balance as
intended:

One interface is at 922mbits outgoing while the other is just at 550 mbits.

Any idea how we could optimize this? How does the loadbalancing algorith on
the mlx work? Is it per-mac-destination or per-flow? To me it looks like per
mac and the algorith seems to be too stupid to move some load from one
interface to the other.

Any idea what to do?

Thanks,
Gunther