Help with core switch redundancy

[mikkydoos]

Hi all,

 

Hope you're safe and well.

 

 

I'm currently doing a network infrastructure upgrade... core switches, hypervisors and upgrade to 10G backbone between two main buildings.

 

Site consists of two main buildings and a few ancillary buildings all connected via 1G fibre from either main building.

Currently have one gigabit core switch (dying) per main building connected via fibre (OM4 x 12 pairs).

 

 

I'm after some suggestions regarding configuring and cabling for core switch redundancy.

 

Core switches I'll probably purchase will be Cisco SX500 series 10G's..... can't afford more!

 

Before I buy I just want to get some opinions on how to cable this up properly.

I'm thinking of stacking two core switches per main building... but I'm a little hesitant on how I should cable up the fibres.

I need to maintain connectivity between main buildings should either of the core switches fail.

 

 

Basic diagram of site and proposed layout......

 

 

 

Edge switches will be stacked to that kind of takes care of them.

 

 

Would appreciate you guys throwing some opinions at me.....

 

 

Cheers in advance. Stay safe

[Davit2005]

If all 4 of the SG500 can become 1 logical stack. I'd aim for a ring topology with min 2 fibres between each building on different routes if possible with the switches in each powered from different UPS or power source if they do not have redundant power. If you have enough fibres would it be worth splitting the Hypervisors between each switch? Bearing in mind these are not modular switches with dual management etc.

 

I've got 4 Aruba 2930f switches at home I don't think it matters which way you stack them but I'd look at possibly 1 > 2 > 3 > 4 and then back to one creating a Ring topology rather than a chain.

 

With the Aruba 2930f you can use any front ports as long as speed of ports is matched with neighbour's. With the 2930f you only get a Ring with 3 or more switches cabled and configured appropriately

Edited February 3, 2021 by Davit2005

[mikkydoos]

creating a Ring topology rather than a chain

 

 

.... as in a Mesh ? Y I was thinking of that.

 

I'm pretty sure they can be stacked together.

 

I'm actually looking at two different models of the same range... SX500-24FT and SX50016FT... same switch.. different number of ports. One pair in each building,

 

 

So to achieve that it would be...

 

 

Copper from Core 1 (B1) ---> Core 2 (B1) ---> Fibre from Core 2 (B1) ---> Core 3 (B2) ---> Copper from Core 3 (B2) ---> Core 4 (B2) ---> Fibre from Core 4 (B2) ---> Core 1 (B1)

 

Edge switch stacks out from both Cores in each building (2 uplinks) per stack, one from each Core.

[Davit2005]

.... as in a Mesh ? Y I was thinking of that.

 

I'm pretty sure they can be stacked together.

 

I'm actually looking at two different models of the same range... SX500-24FT and SX50016FT... same switch.. different number of ports. One pair in each building,

 

 

So to achieve that it would be...

 

 

Copper from Core 1 (B1) ---> Core 2 (B1) ---> Fibre from Core 2 (B1) ---> Core 3 (B2) ---> Copper from Core 3 (B2) ---> Core 4 (B2) ---> Fibre from Core 4 (B2) ---> Core 1 (B1)

 

Edge switch stacks out from both Cores in each building (2 uplinks) per stack, one from each Core.

 

Not really a Mesh but a ring. In a ring you could lose any one switch and it will just effect connectivity to that one switch. A mesh would be 3 connections on each switch so each switch has a connection to every other switch but I am not sure if that is supported in stacking or even needed TBH.

 

You could split each switch to each core in theory.

Edited February 3, 2021 by Davit2005

[mikkydoos]

Not really a Mesh but a ring. In a Ring you could lose any one switch and it will just effect connectivity to that one switch. A mesh would be 3 connections on each switch so each switch has a connection to every other switch but I am not sure if that is supported in stacking or even needed TBH.

 

Yes you are correct @Davit2005.. my bad.

 

Also, as long as I have two uplinks to each edge stack, one from each Core, to edge stack no.1 and stack no.X the the lost of any core switch is then negated.

[Davit2005]

If you have a LAGG group configured or not and just relying on spanning tree to block redundant links it does not matter as the core is acting as one switch and as long as you split between switches you should be fine.

 

Of course if you rely on Spanning tree then one link will be blocked to prevent loops make sure the spanning tree is compatible as well as I have seen issues between devices that have had different Spanning tree protocols enabled.

 

A LAGG link would be preferable as you can utilise both links but suggest checking if a LAGG can be configured between the Cisco switches and your edge switches most of the time it can but may need a bit of tweaking so just check to make sure.

Edited February 3, 2021 by Davit2005