Layer 3 Load Balancing and Redundancy

[gunnydaman]

Hello,

So I am standing up a new Layer 3 LAN with all connections being Layer 3 connections. The goal is to have two layer 3 core switches and several user access switches with a connection back to both cores.

When using layer 2, I know you can use VPC between cores and then use HSRP to point traffic to one Core or the other, and when one Core fails traffic will automatically re-direct to the the second core. However, When it comes to Layer 3, this is not a viable option.

Is there a protocol I can use between the two Layer 3 core switches and the Layer 3 user access switches, which will provide load balancing and redundancy? I was looking at GLBP as a possible solution but I am not sure how the configs would work on the user access switches. I have provided an image to give an idea of what I am talking about. 

Any help would be appreciated.

Thanks,

Matt

[paul driver]

Hello

Are those two L3 Nxos switches going to be in a VPC domain?

Do you require a L2 exteneded vlan design or can the acces-Layer also be L3?

[gunnydaman]

Hello Paul,

There will be no VPC domain involved on the Nxos'. The access layer will also be layer 3 (interfaces with ips and layer 3 interface vlans).

[Joseph W. Doherty]

If your L3 is using a dynamic routing protocol, many will do ECMP load balancing and provide resiliency for your traffic. I.e. I would suggest using a dynamic routing protocol over some combination of static routes and a FHRP. (BTW, for the latter, something like mHSRP would likely be a better choice than GLBP.)

[gunnydaman]

Hello Joseph,

So are you saying that I should use a Dynamic routing protocol between the user access switches and the cores and MHSRP between the two cores?

Thanks,

Matt

[Joseph W. Doherty · Accepted solution]

No, I'm suggesting you use just a dynamic routing protocol or if you use static routes, then use mHSRP as (some) addresses in the static route.

[Richard Burts]

Matt

thanks for the drawing. It clearly shows the access switch as a layer 3 switch with layer 3 links to 2 core switches. Assuming this is correct I would certainly suggest running a dynamic routing protocol (EIGRP, OSPF, etc) between the access switch and each core switch. As @Joseph W. Doherty  points out this should provide equal cost load sharing and would provide failover to the surviving switch if there was a failure on a core switch.