• September 21, 2020, 09:25:26 AM
Welcome, Guest. Please login or register. Registration is free.
Did you miss your activation email?

Author Topic: Square cluster with non-dual homed "ring" network behind it  (Read 4627 times)

0 Members and 1 Guest are viewing this topic.

Offline Johan Witters

  • Sr. Member
  • ****
  • Posts: 252
    • BKM Networks
Hello guys,

I've got a customer with a square clustered core running RSTP throughout the network. All stacks in the offices for client access and server connections are dual homed to 1 of the 2 clusters. Thanks to SMLT this is all running like clockwork..

But the customer also has a large production hall, in which several racks are installed for wireless, cameras and clients etc... The client has installed fiber from each rack to the next forming a "ring", and a fiber to 1 of the 2 server rooms forming a star network.

I should be able to get everything up and running in that way so all switches use the direct connection to the core switch, and put the ring (red in the diagram) in blocking.

Here is where I'm hitting an issue..  as spanning tree is disabled on my SMLT's and IST's, and the production network is not dual homed, I'm forming a loop as soon as I enable more than 1 direct link..

I already though about playing with rstp priority and port priority, but until now I've not been able to get it sorted out.. and because the edge switches are 3500's, SLPP is out of the question also..

Is anyone able to point me in the right direction towards a solution on this one?
Kind regards,

Johan Witters

Network Engineer
BKM NV


Offline Paul L

  • Global Moderator
  • Hero Member
  • *****
  • Posts: 754
    • paulaleroux
    • Paul's Networking blog
Re: Square cluster with non-dual homed "ring" network behind it
« Reply #1 on: June 07, 2013, 11:55:17 PM »
if those were VSP4000s that would be a SPB Engineer's wet dream.
ACSS- Avaya Enterprise Routing Switch  #8

Offline Paul L

  • Global Moderator
  • Hero Member
  • *****
  • Posts: 754
    • paulaleroux
    • Paul's Networking blog
Re: Square cluster with non-dual homed "ring" network behind it
« Reply #2 on: June 08, 2013, 09:52:27 AM »
This comment is not meant to be facetious. But what is the customer's requirement and thinking when they had that ring of fiber installed.

I wouldn't spend time trying to get STP working. I would spent time and effort convincing the customer to just bring everything back to the cores or daisy chain where you can't. 
ACSS- Avaya Enterprise Routing Switch  #8

Offline Telair

  • Global Moderator
  • Hero Member
  • *****
  • Posts: 965
Re: Square cluster with non-dual homed "ring" network behind it
« Reply #3 on: June 11, 2013, 01:10:30 PM »
Unless they are running the SPB system that ring is a screw-up on the client's part.  To work they need links back from each switch or stack of switches to the IST'ed core like the other switches are configured.  A ring topology is not supported.

To salvage things, would building 5500 IST'ed clusters be possible?  Just a thought.

Offline Johan Witters

  • Sr. Member
  • ****
  • Posts: 252
    • BKM Networks
Re: Square cluster with non-dual homed "ring" network behind it
« Reply #4 on: June 11, 2013, 04:39:20 PM »
Apparently the customer now only had a physical star topology installed, which resulted in the need for manual onsite intervention: swapping switches/gbic's and repatching fibers between closets..

They pulled extra fibers between the closets to get a ring throughout the warehouse, in order to get some redundancy. But it appears it was never a requirement during the design phase to dual-home all closets...

To get everything up and running  we decided to split up the big ring into smaller sections, and to terminate the ring on the top cluster node. Each switch directly downstream from the top cluster node has a 2nd link patched to the bottom cluster node. In order to maintain the integrity of the network these links were put "admin down" on the bottom node.. In case of issues the customer only has to enable the backup links of the affected ring on the bottom node.

The rings were not split between the cluster nodes to keep the design simple for the customer and limit the chance of making mistakes..

I'm afraid it is not the best design, but I had some technical limitations:
- dual-homing all closets would result in shortage of both SFP's and VSP's
- edge switches are 3500's, so no possibility for SPB, SLPP etc
- no stacks in the warehouse, so terminating the ring in the warehouse and dual-homing 1 switch to the cluster would result in single point of failure

SPB would have been so great if it had been possible..
Kind regards,

Johan Witters

Network Engineer
BKM NV

Offline TankII

  • Hero Member
  • *****
  • Posts: 556
Re: Square cluster with non-dual homed "ring" network behind it
« Reply #5 on: June 13, 2013, 10:52:03 AM »
Get Index-matching Fibers, LX GBICs, and run the distance as SMLT that way.

TankII

Offline Johan Witters

  • Sr. Member
  • ****
  • Posts: 252
    • BKM Networks
Re: Square cluster with non-dual homed "ring" network behind it
« Reply #6 on: June 13, 2013, 11:38:43 AM »
I've heard of index matching gel for fiber connections, but index-matching fibers...

I'm afraid I'm not following, could you please explain?
Kind regards,

Johan Witters

Network Engineer
BKM NV

Offline TankII

  • Hero Member
  • *****
  • Posts: 556
Re: Square cluster with non-dual homed "ring" network behind it
« Reply #7 on: June 13, 2013, 12:36:48 PM »
We purchase Index-matching fibers so we can run over longer distances using our 62.5 micron infrastructure.  Use LX GBICs and we go as far as 1500 feer, clean.
We buy manufactured cables with the appropriate ends (LC on GBIC, SC/LC/ST on the MM interface side) and this can get us SMLT's into buildings that are well beyond the 990 foot range of MM over 62.5.

You could go to one of the buildings and SMLT back to your core using each side of the fiber 'ring' to get there.

TankII

Offline Johan Witters

  • Sr. Member
  • ****
  • Posts: 252
    • BKM Networks
Re: Square cluster with non-dual homed "ring" network behind it
« Reply #8 on: June 13, 2013, 03:37:30 PM »
ah, like that... we call them "mode conditioning" cables...

The issue is not the range, or the amount of fiber strands. We do not have enough SFPs to dual home all switches to the VSP cluster, and if we would, we would not have enough slots on the VSPs.
Kind regards,

Johan Witters

Network Engineer
BKM NV