Welcome! To hierarchical network design, the Cisco enterprise LAN architecture.
When most people think of hierarchy, you might think of like a government or monarchy or even a business structure or something, right? Where you have a king or CEO, and your… lords, or uhh Managers idk, and your peasants or your workers. Well the concept applies to network design as well.
First up, I wanted to start where most networks would try to become; that’s the collapsed core design. This is great. You have your redundant links to the core-distribution layer so even if one of those guys bursts into flames you’re still in business. This as you might imagine can scale up pretty big, but at a certain point you might feel constricted; usually that happens when you start having multiple buildings. But note that your organization can be rather large, several hundred employees, and still operate just fine on a collapsed core design, there’s nothing saying you absolutely need 3 tiers at a certain size or level, but there will come a point when it’s pretty clear that aggregating the distribution layers in a core layer is a good idea.
At this point, traditionally the hierarchical design model has 3 distinct layers. Core, distribution, and access. Now, you might be looking at this and be like ‘well that just looks like a waste of switches. Well you’re right! The whole point though is it’s hugely scalable. Just take that distribution and access block as a cookie cutter and stamp another one on. Then maybe you bought the building across the street too. So you get a crew to run your 10G fiber under the street and you set up another distribution and access block; or MODULE, to your hierarchy. It really helps to start thinking in terms of distinct little modules or cookie cutter type boxes you can just duplicate where you need that module’s functionality to be scaled out to.
So, the hierarchical design as a whole has some good benefits. First of course is how scalable it is. Next, you may have noticed there’s redundancy all over the place. You can trip over a cable, or lose a whole switch and traffic will keep flowing after reconvergence. It also adds a sense of order to your design. Imagine each of these switches down here is a floor, you could name these and connect them in a manner that makes sense and is easier to understand than a sprawling set of switches.
This also means improved fault isolation and predictability. Now what do I mean by that. Say down here one of your wiring closet switches dies. Who’s affected by this? Well you know for sure that it’s just that floor.
So I’ve been biting my tongue a little bit here to try and not bring up the fact that we have layer 2 switches down here in the access layer. Now this is very traditional because only until recently, layer 3 switches were expensive, right? You might not have had the budget to replace each and every one of your wiring closet switches with a layer 3 switch. What does this really mean, well your layer 2 boundary is up here at the distribution layer, so your VLANs could spread end to end. A lot of people like that, making it so you can have your voice vlan be
the same vlan or subnet throughout your office. One big disadvantage of this though is that you have all these redundant links that aren’t being used because you would have layer 2 loops. This is why spanning tree exists, and with your layer 2 boundary at the distribution then in the event of a failure you end up at the mercy of spanning tree’s convergence time to recover. And sure 20sec for rapid spanning tree doesn’t sound too bad, but in a time when your company lives and breaths on the network connectivity this can be a painstaking period of time. That, and your infrastructure is just underutilized since you have all these links that are just waiting to be used in the event of a failure.