Benefits of a L2 switching core/backbone..

johnnynodough

Discussion -

Besides cost, and perhaps a smidge of speed depending on the devices, what would be an advantage of using a Layer 2 switching backbone? This is popping up in my CBT and I havent got a good explanation of the reasons for a L2 backbone, and havent got me book yet. Would certain situations fit this choice better?

keenon

i can think of wire speed switching/fowarding, low latency, no routing or heavy packet examination, cost, ablity to upgrade bandwidth easily using etherchannelling .. this is just off the top of my head so if i miss somthing it wasn't intentional

o and qos

johnnynodough

Couldnt most of those reasons be overcome with good addressing design, scalable toplogy design, and good equipment choice?

Of course, that implies a good budget as well

johnnynodough

Also if you sacrifice L3 at the Core, arent you going to just push the reduced overhead on the distribution layer anyways?

nick619

L3VPN 2547bis is the way to go

johnnynodough

^

Im failing to follow you....Im just talking about the Core of the network and the choices for L2 or L3 switching.

pizzafart

I think that cost is the only reason. This is significant, as some layer 3 functions can add 25% or more to the cost of a Cisco device.

As far as advantages, L3 allows quicker convergance and load balancing because there's no longer a need for spanning tree. That's a pretty big deal. I don't really know a way in which L2 would be preferable. If someone thinks different, feel free to chime in.

Allan

Humper

Doesn't switching use ASIC hardware chips which is alot faster then software based? : :

wildfire

Wow this is an old topic been opened up again.

Ok ASIC does provide wirespeed switching at L3 But! the destination must already be in the FIB table (or adjacency table) if not the switch will execute a CEF PUNT! this means it uses the L3 engine which is done in software. A number of other reasons cause this, FIB table being full, tunneled or encrypted packets, MTU exceed, ICMP messages, the list goes on.

The core of a large campus network should have one taks in mind move packets as quickly as possible with redundancy. I like to think of it has a freeway or motorway, you (hopefully) can get to you dsestination faster as you dont need to turn off it or make decisions all the time, when you get to your turning you come off onto a smaller road (distrubtion layer) and from there decide which roads to take.

This model is only theoretical though smaller Campus networks use the Collapsed core method where the distro and core layer are combined, so it all depends on the size of your network and your needs.
But remeber the core is simply about speed! all the other stuff is done by someone else (other layers)

DW [banned]

Kewl

Paul Boz

Switching is extremely fast and robust, and can support redundancy much cheaper than layer 3 networking. The core shouldn't be concerned with where data goes, it should be concerned with being wide open so that layer 3 devices can use the "freeway" to get information transferred very quickly. At the core you want as much speed and reliability as possible.