etherchannel load-balancing and unpredictability

[spfister336]

I have a question regarding etherchannel load balancing. I've got a 4507R switch connected to a 3560 switch by means of two content filters which are acting as transparent bridges. The two ports on each side that the content filters are connected to are set up as access ports and are in an etherchannel. The load balancing method on each switch is set to src-dst-ip. I was under the impression that each pair of source and destination ip address would select exactly one content filter no matter which direction.

I've been told that this can be 'unpredictable' and may cause assymetric flows. The algorithm seems fairly straightforward to me. I don't see where the unpredictability can come in. Can someone explain to me what I'm missing?

[Jon Marshall]

removed as need to do a bit more reading....

[Rahul Kachalia]

Hello,

The egress traffic load sharing across each member-links of EtherChannel is directly influenced from the "input variables".

This input variables is either default or user-defined, i.e. src-mac or src-dst-ip.

Based on this input variables each system pre-computes hash and derives some results to program ASICs to load share traffic across each available physical paths in the bundle link. Even when you configure same input variable on each end of EtherChannel interface the result may differ based on way each system performs hash computation, i.e. 3560 hash results may not be same 4500 in your case. So this is something you may call "unpredictable", while it may have symmetric fwding path it may allow you to utilize all available fwding paths from that EtherChannel. This may even help in reducing number of flow count impact when either of the member link failure occurs.

For optimal hash and load sharing results the recommendation is to bundle 2, 4 or 8 physical interfaces in EtherChannel. You may refer to EtherChannel Load-Sharing section in Chapter 2 of following design guide :

Borderless Campus 1.0 Design Guide

thanks,

rahul.

[Nathan Spitzer]

Actually, there is nothing "unpredictable" about the Etherchannel path algorithm. It is 100% deterministic. While the algorithms/hash protocols are proprietary you can GUARANTEE  a flow (as defined by your hashing algorithm) will always take the same path if that path is available. Note that the has algorithms are computed separately by the devices on each end of the link and need not be the same.

It may be that what is being referred to is your content switches not liking that. If I understand your setup the content switches are in the middle of the LACP etherchannel. Honestly, my "knee-jerk" reaction is that that design is just asking for trouble unless you can find documentation where that setup is explicitly supported and even then I would not recommend it.