L3 Etherchannel: benefits?

Deathmage

What would be a use case for L3 Etherchannel? ...IE: maybe Core to Distro links for backbone Area 0?

Mooseboost

This video may help, Jeremy explains it better than I could: https://www.youtube.com/watch?v=8c4sUpZ5adU

TWX

I have several places where multiple discreet campuses adjacent to each other connect through a single service entrance router. In those cases, if I were to want to aggregate links to give them more than 1G I would have to use L3 Etherchannel.

powmia

L2 vs. L3 is one thing. Independent links vs. LAGs is another.

The benefit of L3 is that it's not L2

The benefit of a LAG is that it has more capacity.

late_collision

Deathmage wrote: »

maybe Core to Distro links

^^ This is a good place for EtherChannels. They may also be used between the Distro and Access layer as well as between Distro switches, depending on where your Layer 2 / Layer 3 boundary is.

In addition to increased bandwidth, there are other benefits as well, such as increasing network stability. If you had a link or interface fail, as long as sufficient links remain, the EtherChannel remains active. This prevents a routing protocol convergence / link failover event. Be mindful, if the (now fewer links) EtherChannel becomes saturated, traffic may be lost. You can use the "port-channel min-links <num_links>" command to mitigate this.

Above, I mentioned the increase in stability, to increase redundancy, you may use more advanced designs incorporating VSS or vPC to create mutli-chassis EtherChannels. Here is a link for more information:

Cisco Catalyst 6500 VSS and Cisco Nexus 7000 vPC Interoperability and Best Practices White Paper - Cisco

OfWolfAndMan

Benefit to L3 links is ECMP.

Benefit to etherchannels is increased aggregate throughput, and by aggregating L3 links, you reduce path number by 1, potentially reducing extended reconvergence or recalculation of paths during state change.

Deathmage

OfWolfAndMan wrote: »

Benefit to L3 links is ECMP.

Benefit to etherchannels is increased aggregate throughput, and by aggregating L3 links, you reduce path number by 1, potentially reducing extended reconvergence or recalculation of paths during state change.

ahhh, that's perfect.

I might turn off switchport then on the etherchannels in the home-lab on L3's interconnects with 'no switchport' and assign a ip address to the EC link. I always want the access-layer EC bonds to be in a 'spanning-tree bpdu enable' state on the DP side so that they prefer the distro-layer EC to the root bridge (of course if the distro-link fails it would go the other way), but now if a L3 EC reduces the number by 1 then I can see were by default that value would be lower and the distro-layer would be the preferred route.

Of course I like redundancy upon redundancy so I have criss-crossed links from access to distro so there is no SPOF, really like learning L2 from CCIE level. means when I take the CCNA it will dumbed down

Been doing a ton of 1 on 1 mentoring with Rob, he's a CCIE I meet online after he saw my blog and messaged me on Facebook. I'm like a information sponge when we talk.

networker050184

I never go with L3 etherchannel. ECMP is definitely the way to go there for ease of management and troubleshooting.

Deathmage

I learn something new every day, had to lookup ECMP, interesting.

james43026

Just so everyone understands, ECMP is CEF in the Cisco world, for those who may not believe me, here is an engineer at Cisco explaining this.