Design High Availability with two switch core Summit 460-24P

  • 11 May 2016
  • 8 replies

Hello and good morning
I have a doubt about a design for an small network, consist in two stack one for data and one for voice, those stack are connect it to a switch core, so i want to put another switch core for redundancy and high availability, so i want to know if my design its the correct, can you help me because in my bloody job there is no one to talk about it.

1.- Connect the second switch core to the switch core in production with a etherchannel with 2 ports attached (¿L2 or L3 etherchannel?)

2.-a tier 2 topology will be create it, so from the data stack(3 switch) i connect it to the 2 switch core in in a form of "v shape" or triangule? so 2 ports for each switch to connect to the core?

3.-the same with de voice stack(2 switch) 2 ports for each switch to the core switch,

Can you validate that design please!!!
because i'm 24 and this is my first designing network, so i guess i'm very nervous
Thanks for your time
Best regards!

8 replies

Userlevel 6
Hi Sebastian,

The physical network design would be the following based on your description:

I believe that's a good option. However, to accomplish that you have to decide among couple logical options using protocols like MLAG/VRRP/EAPS/ERPS, etc.
Hi Sebastian,

I agree with Henrique : the dual attachement design is simple and effective.

I would add the simplest solution would be to stack your core switches so they are seen as a single device. You can do this using either stack modules (SummitStack, SummitStack-V80 - in slot B only ! - or SummitStackV over 10Gb uplinks).

The advantage over a simple etherchannel/LACP aggregation (L2 protocols) is that you'd have only one config for both core switches, which means :
- you can aggregate links between the core and acces stacks (using MLAG/LACP)
- you have only one config for both core switches
- you don't need to worry about L3 redundancy (no VRRP)

Anyway, it will be the easy way, compared to a triangle architecture (loop prevention with SpanningTree or EAPS, routing redundancy with VRRP...)
Hello Henrique,

Thanks for all your help, but in the MLAG from the switch that are connect with LAG, can i do LAG from an example the Data Stack to the core2 ?

or the MLAG feature is asymmetrical from a point of view, does MLAG required that boths ends have to be in stack?
or just one side?

Userlevel 6
Hi Sebastian,

MLAG will be created between the 2 Core switches. The links between them (with LAG enabled) will be ISC link used to sync fdb table entries.

Regarding the stacks you just need to enable sharing (LAG) for the 2 ports connecting both Core switches.

The vlans from stacked switches must be added to the Core ISC link as well as its MLAG ports (the ports in the Core that connects the stack).
Hello Henrique
thanks for your help Enrique, i'm going to do it, but it give me a doubt, because i'm going to connect it in triangule shape, looking the vlans configuration that are going to pass, which are the Data and Voice,

the vlan configuration says that is not running STPD, and i learned networking from other vendor, but that a vlan i not running in a STPD instance is not good,

Admin State: Enabled Tagging: 802.1Q Tag 6
Virtual router: VR-Default
IPv4 Forwarding: Disabled
Primary IP :
IPv6 Forwarding: Disabled
IPv6: None
STPD: None

Looking in the Forum, i supposed, that when a create a vlan i have to create an STPD instance?

configure stpd s0 mode dot1w
conf stpd "s0" add vlan "Default" ports all
conf stpd "s0" add vlan "Voice" ports all

Because in other post it show this
with this configuration the switches assumes that all the ports assigned to vlans are in the STP instance?

Userlevel 3
Hi Sebastian,

If you are using MLAG on the core switches to connect to the voice and data stacks, this will prevent loops when it's enabled and therefore STP is not needed. You do not need to enable STP when you create VLANs on EXOS switches.
Hope this helps.

Hi Tone,

Right now the data stack are connect to one core only because i have no stack connection between them, because right know i don't have the modules and i don't know when arrives,

So my doubt is this, in the switch core2 i have lag connection with the data stack, so what happens when i connect the data stack to core1

Both core1 and core2 are connect between them with a LAG, so if a create another LAG between data stack and Core1, doesn't create loop?
because i learned with cisco, so thats my doubt, because i thinks it's going to make a loop

data stack --> core 2 --> core2 to core1 --> redundant connection between Core1 and the data Stack?

Does extreme have something enable for blocking redundant ports? because the Data vlan show that has not STP

Userlevel 3
Hi Sebastian,

You can implement MLAG with the 2 core switches as MLAG peers and the data stack connected to both peers. There will be a physical loop, but MLAG will block packets going across the ISC link to prevent a logical loop. Below is some information and a configuration example that may help.

How to configure MLAG in Extreme switches

Hope this helps.