• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Routing at the distribution layer vs core

her209

No Lifer
If you have hundreds of VLANs and you have L3 switches installed in the distribution layer, would it be preferable to do the routing at the distribution layer and leave the core as a pure switch. Reason being, the process of routing is distributed to more, albeit, slower switches which is better because routing is more processor intensive than switching especially when talking about millions of packets per second.
 
Most L3 switches route at line speed, so speed isn't an issue.

However, because of advances in network technology, as well as decreases in pricing, most organizations favor a collapsed core (distribution and core are effectively the same layer) over a three tier design. This simplifies the topology quite a bit. Optics are cheaper and fiber is faster than it ever had been before, so the three tier design isn't really necessary anymore within a corporate network.

Also, it's less about the number of VLANs as it is about the number of switches.

But, because L3 switches are so fast and so inexpensive now, there's no reason not to use them in the access layer.

Really, it's all about what you want to accomplish. If you need a single VLAN to span multiple access nodes (which is really bad design by itself), you'll need a three-tier design. Layer 2 traffic should never, ever cross your core.
 
Basically what Drebo said. My goal is to allow packets to take the least number of hops so using a L3 switch out at or close the access layer is what I use. I don't use L3 on all the access ports but if I have a cluster of L2 switches that handle a floor or 3 of a building those will link into an L3 switch before going back to a farther up the chain IDF. I want any intra ip routing happening on the 3 floors to stay out at the far L3 gear if it can. I don't want the pbx [computer] application needing to connect to a phone on the voice vlan via the core when it should be local[ish].

Also I agree that L2 should not be crossing core but periodically you do run in to crappy company applications that need that for some stupid reason and it so happens the CFO and accounting are in differing buildings / floors, you can still bring VLANs out across the core if needed. It just adds more work and should be the exception, no the rule.

--edit--

You just need to be comfortable with dynamic routing protocols to make this work like a top.
 
Last edited:
I agree w/ drebo that L3 switches nowadays route traffic in hardware, so performance difference is really minimal, although the route learning itself is still a CPU bound function.

The collapsed core design makes sense for small/medium network, but for large networks I personally still prefer three tier topology. Exception is if you have a lot of east/west traffic pattern in a datacenter, then a different approach can be taken, such as FabricPath/QFabric.

10 years ago the slogan was "switch wherever you can, route if you must".
Nowadays it's the other way around...routed links converge so much faster, and it's a lot easier to manipulate routed traffic, w/ a lot of flexibilities.
 
Back
Top