Link aggregation combines multiple physical network links between switches into a single logical connection, increasing total bandwidth and providing redundancy. Configuration requirements, load sharing behavior, and key differences from Spanning Tree Protocol are also covered.
Link Aggregation
Link aggregation is when we combine multiple links together to create more bandwidth between two switches. As an example, I might have two switches that have ports on them that operate at 100 megabits per second. If I were to just connect one of those links together, then perhaps that's not enough bandwidth between those two switches to facilitate all my traffic. I may need more than 100 megabits per second between those two switches. So what I can do is plug in multiple ports — for this example, say four of those ports — and aggregate them together, and now I have one link between those switches that operates at 400 megabits per second.
We talked about spanning tree protocol, and during that process we actually created multiple links between our switches. We created two links between every single switch to create redundancy. But we also talked a lot about spanning tree protocol and how spanning tree protocol allows this to happen, because it eliminates or shuts down some of these extra links. It will turn off some of these links so they're not all operating, and that mitigates against any kind of loops.
Well, link aggregation allows us to actually utilize those links. What I can do is combine these two links over here into one link, therefore increasing the amount of bandwidth that I have across there. And I can do it for two, three, four, five, up to eight different links I can band together. So if these are 100 megabit per second connections, then I can band up to eight of them together and get 800 megabits per second between these switches. If they're gigabit connections, I can band up to eight of them and get eight gigabits per second between these switches. So link aggregation is great, and it also allows for that redundancy, because if one of those links dies then the connection still remains up and running.
Now, link aggregation goes by many different names. Link aggregation is the most common, but you'll also hear the terms bundling, bonding, channeling and teaming. Cisco calls it EtherChannel, and we'll talk more about EtherChannels and what EtherChannels look like. Midway through this module we're actually going to switch our terminology and just start talking about EtherChannels, but I'm going to talk about link aggregation right now because link aggregation is not Cisco specific.
I like to relate it to spanning tree protocol because there are some similarities, there are some overlaps between link aggregation and spanning tree protocol. But, once again, they serve two very different purposes. Spanning tree protocol actively looks for redundant links and shuts them down, and allows you to mitigate against layer 2 loops that happen and broadcast storms that happen. Link aggregation allows us to bundle those links together to create more bandwidth between our switches.
First of all, link aggregation creates redundancy between two switches. I will bond the links between two switches together to create some redundancy within there, so if any one of those links fails it's still up and running. Spanning tree protocol allows us to create redundancy between many different switches. Link aggregation doesn't do that — I mean, we can have links between many different switches, but they can't all be bonded together into one aggregation.
The other thing that link aggregation allows us to do is some load sharing amongst those different ethernet cables. There are some different ways that we can load balance, or we can share the load between those different links. It's kind of interesting how some of that does that, and it's beyond the scope of this course, but just know that link aggregation really just shares the load amongst all of those bonded links.
Spanning tree protocol doesn't really have that capability. They don't really have load sharing amongst the multiple links. The one exception to that is that we can actually do a per VLAN spanning tree protocol, so if we wanted to engineer our network so some VLANs use one link and other VLANs use other links, we can actually divide that up and have some load sharing. But it doesn't dynamically change. The load sharing is just based per VLAN, and you could have one VLAN that just needs a ton of data where another VLAN doesn't need as much bandwidth between those switches, and spanning tree protocol is not going to adjust for that, versus link aggregation will. We can utilize all of those links.
Link aggregation increases the bandwidth of a link, and once again spanning tree protocol has just a dedicated link per VLAN, very similar to what I just mentioned.
Link aggregation does not eliminate loops that happen within your network. It can eliminate the need for spanning tree protocol just between the links that are aggregated, but it doesn't involve other links and other connections throughout your network, and where you want multiple backups to our backups. So with all of that we still need spanning tree protocol, because it really eliminates the loops in your network. So those are some of the similarities and differences between link aggregation and spanning tree protocol.
You can have anywhere from two to eight of these links bonded together to combine their bandwidth, but there are some configuration guidelines along with this.
First of all, what you should know is they don't have to be physically contiguous ports. What I mean by that is they don't have to be right next to each other on the switch. This port could be 0/1 and this port could be 0/48, and I can still link them up. Now, generally speaking we do have them right next to each other on the switch. I like to usually put mine right towards the end of the switch and have them grouped up all together — 0/47 and 48, or 0/45, 46, 47 and 48 — and group them up that way. It just makes more sense when you're looking at the configuration, and it makes more sense when you're plugging in, but it doesn't have to be set up that way.
The other thing is that you must have your configurations match on both sides, and there are several ways that they must match.
If you are configuring these, you can still configure the individual ports — in fact we're going to have to do that to bring them together onto the same link aggregation — but then I recommend configuring the actual link aggregation as a whole, because if you put two different configurations on each one of these ports then it will not function correctly. Those configurations need to match.
Some limitations of link aggregation: you can only bind up to eight connections at any given time, so that's one of the limitations. Another limitation is you can't combine a Fast Ethernet and a Gigabit Ethernet at the same time, so this side can't be at a hundred Meg and this side at one gigabit per second. That type of connection isn't going to be allowed. They all have to be 100 Meg, or they all have to be a gigabit connection. Also, there's a limit on how many EtherChannels you can have on a single switch, or link aggregations on a single switch. For instance, a 2960 can only have up to six of these EtherChannels on a switch. So there are some limitations on how many of these EtherChannels, or how many of these link aggregations, I create.
TechKnowSurge builds IT and cybersecurity professionals through hands-on, concept-first training built around real understanding — not memorization. Free interactive tools, structured programs, and 25+ years of real-world experience, all in one place.
Explore free tools and programs →