Network LAN Design and STP

Answered Question
Jan 29th, 2010

Hello all,

I've recently acquired a couple of Catalyst 3560 10/100/1000 switches and am trying to implement some layer 2 redundancy with STP.

I've uploaded a document showing 5 switches, using only 3 vlans.

I would like to know if this implementation of STP is correct, and if it allows for optimal redundancy.

Please feel free to point out any errors, or suggestions on how I could improve my topology.

These switches will eventually go into a production network, and I want to make sure I have it right.

I have this problem too.
0 votes
Correct Answer by Jon Marshall about 6 years 10 months ago

nelson.garcia wrote:

Hello all,

I've recently acquired a couple of Catalyst 3560 10/100/1000 switches and am trying to implement some layer 2 redundancy with STP.

I've uploaded a document showing 5 switches, using only 3 vlans.

I would like to know if this implementation of STP is correct, and if it allows for optimal redundancy.

Please feel free to point out any errors, or suggestions on how I could improve my topology.

These switches will eventually go into a production network, and I want to make sure I have it right.

Nelson

Personally i think there are too many interconnections here and too many squares as opposed to triangles. What i would do is to make both 48 3560 switches STP root and secondary. Then connect each 24 port 3560 to both 48 port switches. There is no need to interconnect the 24 port switches and it just adds complexity without adding any extra redundancy.

Setting it up this way will also make it easier to troubleshoot.

Jon

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (3 ratings)
Loading.
Correct Answer
Jon Marshall Fri, 01/29/2010 - 09:16

nelson.garcia wrote:

Hello all,

I've recently acquired a couple of Catalyst 3560 10/100/1000 switches and am trying to implement some layer 2 redundancy with STP.

I've uploaded a document showing 5 switches, using only 3 vlans.

I would like to know if this implementation of STP is correct, and if it allows for optimal redundancy.

Please feel free to point out any errors, or suggestions on how I could improve my topology.

These switches will eventually go into a production network, and I want to make sure I have it right.

Nelson

Personally i think there are too many interconnections here and too many squares as opposed to triangles. What i would do is to make both 48 3560 switches STP root and secondary. Then connect each 24 port 3560 to both 48 port switches. There is no need to interconnect the 24 port switches and it just adds complexity without adding any extra redundancy.

Setting it up this way will also make it easier to troubleshoot.

Jon

nelson.garcia Fri, 01/29/2010 - 09:30

Jon, after drawing up your diagram I did notice that it is indeed less complex.

I compared my diagram to yours and saw that it was almost similar, except that the 24 ports had interconnections to each other as well, maybe that was over doing it a bit?


Anyway, thank you for the response, Jon.

One thing though, you mentioned that I had too many squares and no triangles... why does the shape of the interconnection matter?

sachinraja Fri, 01/29/2010 - 09:40

Hi Nelson

what Jon was mentioning is to have a dual core with edge switches on L2, as shown in the diagram.. I made up a very quick basic diagram of what Jon meant.. might be useful for you..

this would have less complexity as Jon mentioned.. it is also a very reliable and highly available network with no single points of failures.. spanning tree configuration is easy this way...

Hope this helps..all the best

Raj

Attachment: 
nelson.garcia Fri, 01/29/2010 - 09:45

Raj, would a connection between the two 48 ports be necessary for any reason?

sachinraja Fri, 01/29/2010 - 10:06

Nelson

Im really sorry.. I made this diagram in 48 secs, and forgot adding the link between the core switches .. Yeah.. you should have a link between core 1 and core 2.. probably a layer 2 link so that the same vlans can be propagated onto both the switches.. yo u can also have HSRP configured between core 1 and core 2, to make sure layer 3 redundancy works..  you need to have a link just to maintain high availability and having consistent spanning tree and vlan information across the network ...

Hope this helps.. all the best

Raj

Jon Marshall Fri, 01/29/2010 - 09:39

nelson.garcia wrote:

Jon, after drawing up your diagram I did notice that it is indeed less complex.

I compared my diagram to yours and saw that it was almost similar, except that the 24 ports had interconnections to each other as well, maybe that was over doing it a bit?


Anyway, thank you for the response, Jon.

One thing though, you mentioned that I had too many squares and no triangles... why does the shape of the interconnection matter?

Nelson

Squares vs triangles is a Cisco design recommendation. Think of it like this -

scenario1 - you have one of your 24 port switches connected to both 48 ports switches and this forms a triangle

scenario 2 - you have 2 24 port switches. One is connected to one of the 48 port switches (SW1)  The other is connected to the other 48 port switch (SW2). And they are both connected to each other. This forms a square.

Now in scenario 1 if either of the 48 port switches fails the 24 port switch will know immediately that it has failed because it has an uplink to it ie. it can detect the failure at a physical level.

In scenario 2 SW2 fails. Now the 24 port switch connected to SW2 knows straight away but the other24 port switch doesn't. It doesn't have a connection to SW2, it is connected to SW1 and the other 24 port switch so as far as it is concerned everything is okay in the network.

That is why Cisco recommend triangles because a switch failure is detectable at the physical layer rather than having to wait for certain timers to fail. This is applicable to both L2 and L3 connections.

Hope i've explained that clearly enough !

Jon

Actions

This Discussion