Question on STP redundancy

Unanswered Question
Sep 3rd, 2008


Besides loop avoidance, does STP (or newer instances like RSTP or MST) provide redundancy in case some of the links/ports fails like routing protocols do?



I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Jon Marshall Wed, 09/03/2008 - 03:56


That is one of the key features of STP in that you can design networks without any L2 loops in them, in which case you would not need STP, but more often than not switches have redundant connections because as you say some links, ports can fail.

Without STP these redundant links often form a loop and so STP blocks on one of the links to prevent the loop. If the active link fails STP can then unblock this port without creating a L2 loop.

Note that another way of providing port failure is to use etherchannel where multiple ports are aggregated to single link and STP see the single link only so it does not block any of the individual ports.


markopstc Wed, 09/03/2008 - 07:47


Thanks for the reply. At the moment, I'm looking at the best implementation for switching core and one of the options is STP, while another is some of the link-state routing protocols (all switches are L3 capable).

Can tou advise on pros/cons for both of these?



francisco_1 Wed, 09/03/2008 - 08:05


One of the design i did recently, i used 2 6500's as core switches connected via L3 and L2 (trunks) to the access layer switches Cat 3750. The Design was based on Core/Access layer no distribution. On the access layer, No VLAN's spanning across multiple access layer switches so no need for STP between the access/core layer and no STP between the cores since i used layer 3. Traffic between the vlans is routed via the core switches. With my design, i have no concern for STP convergence. The recommended design for hierarchical layer is if you can do without STP, then don't use it if you don't have any requirements for it. I have worked in environments where they are routing on the access layer using EIGRP instead of using STP.

see this doc for better explanation including pros/cons when using STP or Dynamic routing.


Joseph W. Doherty Wed, 09/03/2008 - 16:24

One major difference is, STP blocks all but one path, routing leaves them open, and depending on protocol and topology, might use multiple paths. Another difference is speed of convergence, which could be epecially slow with the original STP, but later variants are usually much faster.

Modern designs tend to recommend L3, since modern L3 switches are usually nearly as fast as L2 switches.

markopstc Thu, 09/04/2008 - 01:30

Thank you all for replies, much appreciated.

Since L3 is recommended in the core, would you recommend MPLS as layer 2.5 with OSPF as the best choice?



Joseph W. Doherty Thu, 09/04/2008 - 03:35

I haven't worked with LAN MPLS. However, seems a big jump from from either a typical LAN's L2 or L3 unless you really need multiple VRFs within your LAN. Perhaps other posters might comment on using MPLS within a LAN.

Jon Marshall Thu, 09/04/2008 - 07:11


I agree with Joseph. MPLS in a LAN environment is not necessarily wrong but it really does depend on your requirements. If you have a need to completely separate different end user traffic with separate routing tables etc. or perhaps you have overlapping addressing problems then just maybe MPLS could be used but in an ethernet LAN for an Enterprise it would be a difficult one to justify.


paul.matthews Thu, 09/04/2008 - 11:16

I'm with Jon and Joseph. I really believe in KISS (Keep it Simple...)

MPLS is almost certainly OTT for a lan environment, unless the LAN environment is a multi client shared infrastructure datacentre!

Keep it simple, ad when it breaks at 3AM on sunday morning, and whoever is on shift can sort it. Make a complex solution and what is gonna happen? Yup - you get the call.

As for L2 Vs L3 - with L3 switching being commonplace there is no performance benefit in an L2 solution. L2 will only allow a single path to be active, where L3 can permit multiple paths to be active, and is less prone to things like undrectional link errors. If you were to use something like OSPF with BFD, you could probably knock spots of the STP convergence time as well.


This Discussion