2

I have a topology in which there is a main ring and some nodes in the main ring have a minor ring. enter image description here 1- which STP configuration has the fastest convergence time? How would you configure spanning tree for all switches? I think RSTP does the best, but Is there any other configuration to help convergence time?

2 - Does RSTP have any limitation? limitation In number of switches for instance? or incompatibility with other vendor's RSTP or spanning tree algorithm?

3- Is RSTP practical solution for this topology with a large number of switches? 45 switches as the main ring and more than 100 switches in minor rings?

4- Is there any other faster way, rather than spanning tree, to provide protection for this large network? 

Thank you all! 

Improve this question
1
  • 7
    That is an absolutely horrible topology. It needs to be broken up with routers into multiple LANs. Implementing something like that will probably get you fired.
    – Ron Maupin ♦
    Feb 20, 2019 at 18:10

2 Answers 2

Reset to default
9
  1. MSTP and RSTP converge equally fast, MSTP is the current protocol version. Given the topology in your diagram (ugh!) and its tree depth, no STP variant will converge quickly (if at all).
  2. xSTP default parameters are designed for a maximum tree depth or chain length of 20 hops. Building longer/deeper constructs requires parameter tuning or your network will never converge. Usually, trees are much more shallow with a maximum of four or five L2 levels, ideally only two or three.
  3. Generally, yes. You'll need to organize them better however. Don't build rings (at all). The best design approach in regard of performance, scalability and resiliance is two core switches, a distribution layer to increase the port count for the access layer, and an access layer where your nodes connect. When properly combining L2 and L3 connections, that approach scales up to tens of thousands of ports. At least the connections between core and distribution are redundant, often the distribution-access interlinks as well.
  4. That is a question too broad to answer. You can build a network in more than one way. xSTP is probably the easiest, less L2 and more L3 links scale better.

Edit:

the topology is not something that I can change, All the wirings have been done yet.

In case you've got multiple wires between closets you can still improve the topology significantly. Patching connections "through" = across intermediate enclosures allows you to reduce the tree depth significantly.

With just four links between closets (which is the absolute minimum we're deploying) you could e.g. elect Switch4 and Switch7(4) as "core" switches for separate L3 domains (you'll need L3 switches though). Make them redundant (stack) and connect all ring switches to the nearest stack. Run double L2 links between the domains and your network has significant improvements - it'll run xSTP without (too much) tweaking, has very few single point of failures, and increased throughput (depending on the exact details).

Make sure you run MSTP in (at least) two instances and configure the "core" switches as root bridges.

Improve this answer
6
  • 1
    "Given the topology in your diagram (ugh!)..." That is a polite way to put it. :)
    – Ron Maupin ♦
    Feb 20, 2019 at 19:34
  • Thank you Zac67 for your response! Actually, the topology is not something that I can change, All the wirings have been don yet. What do you suggest to do for configurations? Is it good to put routers in two nodes in the main ring and run VRRP on them?(leading to divide the main ring into 2 LANs) or Is it better to have a router in each node of the main ring? or maybe something else? I appreciate your help with my problem.
    – A.A
    Feb 21, 2019 at 6:01
  • 8
    The topology should change - right now it's a bad design, no matter what you do with it. There is no redundancy beyond the ring and many single points of failure. Converting the branching nodes to L3 is a (somewhat) usable approach to solve the L2 tree depth problem. VRRP doesn't do much good since potentially there are no two redundant routers (unless you double some nodes or run redundant links). You'll need a quickly converging routing protocol instead, like OSPF.
    – Zac67 ♦
    Feb 21, 2019 at 7:31
  • @Zac67,you're right. I admit hierarchical topology is the best approach for implementing a network. (by routers and switches) But why ring topology is prevalent among metro ethernet networks? How metro ethernet rings have protection? Are they TDM-based? (like SONET)
    – A.A
    Mar 12, 2019 at 12:41
  • 1
    @AA Metro Ethernet is Ethernet with some extensions - no SONET, SDH or similar. Usually, these rings are double rings with redundant nodes (switches or routers) which is a completely different scenario (esp. when implemented in L3). As Ron's answer points out, Metro Ethernet can handle rings much better.
    – Zac67 ♦
    Mar 12, 2019 at 13:01
7

If you can use Cisco Metro Ethernet switches, you might consider Resilient Ethernet Protocol, which is designed for ring topologies.

REP is a protocol used in order to replace the Spanning Tree Protocol (STP) in some specific Layer 2 network designs. ... Benefits Here are some of the benefits of REP:

REP offers these convergence times:

  • 3750ME converges between 20ms and 79ms
  • ME3400 converges between 40ms and 70ms
  • Works on existing hardware
  • Predictable, blocked ports
  • Easy configuration
Improve this answer

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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