Any way not to use any router or layer 3 switch in a campus LAN?

[azi123]

Hi everybody!

As I understood, switches can form and organize their MAC address table if the connections between sources and destinations would be bidirectional and have acknowledgment ( Is it true?), so what will happen if there is no router or L3 switch in a relatively large network?  what problems may occur if the network just does switching based on MAC addresses? suppose one subnet for the whole network.

Best,
Amir

[paul driver]

Hello

---

@azi123 wrote:

Hi everybody!

As I understood, switches can form and organize their MAC address table if the connections between sources and destinations would be bidirectional and have acknowledgment ( Is it true?), so what will happen if there is no router or L3 switch in a relatively large network?  what problems may occur if the network just does switching based on MAC addresses? suppose one subnet for the whole network.

Best,
Amir

---

You would have one very large broadcast domain so when a host isn't aware of another host mac-address it will broadcast for a reply thus every node on this large broadcast domain apart from the port the broadcast originated from will revive a copy of broadcast and in a very large flat network when multiple hosts are broadcasting  can saturate it to an extent no one would be able to work effectively, This is why vlans are a good thing as they each create their own broadcast domain so not to allow broadcasts from one vlan to reach a another

Hence when you wish for these vlans to communicate they would use a router or l3 interface on a switch to route the vlan traffic.

[dbeattie]

Hi Amir,

The first thing that you are likely to find is a relatively large amount of broadcast traffic as ARP becomes and issue. I have seen this on a single segment with over 1500 endpoints. Please consider that all devices have to process all broadcasts, so this ends up with a lot of CPU cycles lost. The first place that it became evident was on local domain controller servers, which connected to all PCs. On these servers, there was a lot of churn on the ARP tables.

Perhaps more concerning, but less evident is the inherent risks of a segment that size, which is effectively a single failure domain. A single failure can easily wipe out the whole segment. At the very least it could introduce instability within various protocols including Spanning Tree.

Hope this helps,

Dave

[Joseph W. Doherty]

As the others have noted, the potential issues is broadcasts.

To put it another way, broadcasts change your switched network to a shared media network. I.e. every broadcast packet is sent to every active port (except, of course, the one that generated it).

Then, as also noted by Dave, the host must actually process the frame to determine whether it needs to do something with it. Understand, for unicast and mulitcast traffic, a NIC will not send the frame up the stack to the further process it if it's known to be of no interest. That filtering cannot be done for a broadcast frame.

Lastly, you're creating an ideal environment for some DoS attacks. A simple example would be to send out a broadcast ping.