I've got a quick question and I'd be interested in getting the communities feedback.
If you are designing a solution which involves multiple Layer 3 links between devices would you typically go for a Layer 3 port channel or multiple layer 3 links and use the routing protocol to do equal cost multipath / load balancing?
What's your approach? I can't think that there's a massive difference, a port channel seems easier from a trouble shooting and routing protocol overhead and topology perspective but I'd been keen to understand other people's views.
The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.
In no event shall Author be liable for any damages wha2tsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.
I would use the port-channel, because it probably offers more load balancing hashing choices, additional link routes are not passed into the routing topology and possibly the routing topology is not impacted with the addition or lost of a link.
I concur with Joseph.
Let me add that having multiple ECMP links instead of a single Layer3 EtherChannel would also require you to run the routing protocol over all those links, and you would end up with multiple OSPF/EIGRP/IS-IS adjacencies between the two devices. You would end up synchronizing the working databases of these routing protocols through each link independently, every update would need to be advertised and acknowledged over every link, adjacencies would need to be maintained over each link - clearly a waste of effort.
In addition, note that every network learned through multiple equal-cost links will have its "next hop" field expanded into a list of next hops, one for each link, thus consuming more memory and hardware resources in a Layer3 switch. If you advertised many networks over these links, each of these networks would have the "next hop" field expanded. Even though CEF should reuse the same list of next hops for all routes that resolve to it, the routing table itself (the prototype data store from which CEF FIB is built) will possibly grow in its memory footprint.
All these drawbacks are alleviated by using a Layer3 EtherChannel that appears to be a single link, and thus a single adjacency and a single next hop, to the interconnected hosts.
Thanks both, it's great to get the feedback and point of view from other people in the community.
That makes sense what you've said.