[c-nsp] mpls and etherchannel

[Lukas Tribus]

> Hi,
>
> I have 2x 200mbps microwave links connecting switch 'a' to switch
> 'b'. I have a port-channel configured and I currently get an acceptable
> split of traffic based on source / destination address mac hash.
>
> I am now looking to deploy mpls in this portion of the network and
> I am concerned that I will no longer get much of any traffic split since
> the mac address of the frames will largely be the mpls sender / receiver
> and not much of any diversity here.

So currently you load-balance based on src and destination mac addresses,
and it works, because you have a huge layer 2 network, correct (each customer
is a MAC address on the network, probably)?



> I have read that the solutions would be ecmp

Wrong. ECMP vs port-channeling is not the question here. Both ECMP and
port-channeling will use some underlying loadbalancing code, that is platform
dependent. Whether you use ECMP or port-channeling will therefor most likely
be irrelevant.

If we would know the platform, we could tell you more, but generically
speaking, ECMP doesn't behave any better or worse than port-channeling.



> but then I am worried about the possibility of packet re-ordering...?

packet reordering is a problem only if you do per packet load balancing.

Which you should not do. Really don't. And again the actual load balancing
doesn't really depend on a control plane question like ECMP vs LACP.



> The links themselves are direct point to point and I
> don't anticipate that one would be any faster or slower than the other

Stop it. Don't do per packet load-balancing. Repeat after me:
I won't per packet-load balance.



> I just don't know what to expect or if there is something else I would
> look at for this situation?

Clarify:
- what platforms are you using (the LSR's doing ECMP or port-channeling
  on those 2 wireless bridges)?
- what will your traffic look like once on mpls (unlabeled IP, vpnv4 or l2vpn)?


Lukas