etherchannel load balancing

[Dennis Olvany]

Why does Cisco not offer round-robin load balancing for etherchannel?

I understand that round-robin can result in out-of-order packet delivery, but tcp is well equipped to deal with this issue. Per-packet load balancing is achievable at layer 3, so why not layer 2? Round-robin would provide a far better alternative to packet discards in a layer 2 implementation.

[Peter Paluch]

Hi Dennis,

All hosts and intermediate links on a VLAN would have to migrate to 
jumbo, which does nothing for pps to/from remote hosts. It does, 
however, reduce pps within a given VLAN and between jumbo VLANs which 
will serve to reduce overall pps. Path MTU Discovery should alleviate 
fragmentation issues.

Yes, I am sure it would. I guess your mileage would vary depending on whether the bulk of the high-load bandwidth remains in the same VLAN and thus can take the advantage of larger MTU, or is routed through other networks where possibly the MTU may be clamped down to 1500B again, thereby forcing your hosts to revert back to classical MTU sizes.

Dennis, are you trying to solve a particular problem? You have started this discussion with a general question but I have a feeling that you are seeking solution to a particular issue.

Best regards,

Peter

[Dennis Olvany]

I am receiving traffic from a number of gig links that is then transmitted across a multi-gig etherchannel. Due to insufficient load balancing, the pps received is greater than the etherchannel can transmit and discards are resulting.

[Peter Paluch]

Hello Dennis,

What is the nature of the flow, and what switch platforms are you using? Which load balancing method have you selected? Can you be more specific?

Best regards,

Peter

[Dennis Olvany]

It's a 3560 using src-dst-ip.

[Peter Paluch]

Hello Dennis,

I see. Well, on a 3560, there is no more granular method available, sadly.

I have a feeling we are not moving forward. Perhaps we have reached a point of fundamental difference in understanding the EtherChannel and its optimal deployment.

My personal opinion is that there are good reasons not to have the round-robin EtherChannel load balancing algorithm in Catalyst platforms, and I can understand the rationale behind it. On the other hand, I can see that the available load balancing methods do not produce satisfactory results for you, and I completely understand your disappoinment. Perhaps Cisco indeed should support the round-robin load balancing method with an explicit warning that this load balancing method is in violation of the existing standards, and the use is solely on the user's own risk.

Best regards,

Peter

[Dennis Olvany]

I appreciate your input on this matter. It is certainly useful to explore the various perspectives.