I am working on designing the logical topology for a new data center rollout. I'm reading "Cisco Data Center Infrastructure 2.5 Design Guide" and going with the multi-tier design. There are multiple options for this design, including looped layer 2 and loop-free.

The benefit of the looped design, it says, is layer 2 adjacency for servers that require it. The loop-free topology apparently limits layer 2 adjacency "to a single pair of access switches".

I don't quite understand what this means. I would assume it is referring to how many layer 2 "hops" a server is away from another, but in both cases it seems traffic requires a trip up to the aggregation layer (assuming a L2/L3 switch) to cross vlans. If we're talking the same vlan, well looped seems even worse because servers on the same vlan have to traverse up to the agg layer to communicate versus what seems like a trip directly to the connected switch with the loop free example.

Can anyone shed some light on my misunderstanding?

  • I think I got my loop and loop-free examples backwards in the last paragraph. I meant to say that in a loop-free environment traffic between servers on the same vlan have to travel up to the agg layer while in a looped environment the two switches are directly connected (via trunk) and the traffic simply hops right on over. But, I think the question has been answered anyway. :) – S.C. Nov 13 '12 at 15:52

Layer 2 Adjacency, in Ethernet networks, refers to the idea that a packet send out on a segment can directly reach its destination without traveling through a device which would change the packet.

A simple example:
You've got two computers with a wire in-between; whatever comes out of one computer is essentially guaranteed to be receivable by the other computer. These devices have network adjacency.

More complicated:
You've got two computers with a L3 switch in between, both are assigned the same vLAN. Here again what one computer sends the other will receive without question.

Breaking adjacency
Two computers on the L3 switch again, but one is connected to a port assigned to a vLAN, the other is assigned to a trunk port. Now packets from the first computer will arrive to the second computer with a vLAN tag (assuming a typical setup here...)

Who cares? Well routing protocols do (they usually need to be able to figure out the network topology, and that discovery can be broken by non-adjacency), as well as a myriad of non-IP protocols. It's not all that common for these issues to come up on servers in a DC, but is certainly possible.


We use, well require Layer 2 Adjacency for our VMmotion and Microsoft applications. In your above examples the one with the L3 switch in between would actually break the L2 Adjacency. The way it was explained to us and which some of our L2 protocols need to work means the following. L2 Adjacency means the devices communicating must be on the same subnet. Example: needs to communicate with its MNGT server at This would work because its not traversing any L3 devices such as a Router or L3 switch. If you have 2 different subnets your packet with have to hit the Router or L3 switch to view the routing table and decide where to send the packet. In this case your packet would be changed. The Source & Destination IPs would stay the same but as it travels over L3 your notice your packets L2 information will be the routing protocol such as HDLC. So I would answer the question by saying L2 Adjacency means the devices need to be on the same Local LAN.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.