SDA Deployment - IP Pools Confusion for Underlay

[thenetadmin]

We’re in the process of designing Cisco SD-Access, and I need some guidance regarding IP Pools

Setup

• Building A & B:
  • 2 Border Nodes/Control Planes (BN/CP)
  • 4 Intermediate Nodes: 2 in Building A, 2 in Building B
  • Approximately 20 switch stacks of edge nodes in each building.

Questions:

1. What are the IP pool requirements for LAN automation, underlay point-to-point (P2P) links, and loopbacks?
2. Does the underlay IP pool have to be different from the loopback IP range?
3. Should the underlay IP range between the two border nodes be different?
4. Should the underlay IP range between the border nodes and intermediate nodes be different?
5. Should the underlay IP range between the border nodes and edge nodes be different?

I want to ensure our design is clean and efficient without running into issues later during automation or integration. Any advice or best practices would be greatly appreciated.

[Torbjørn]

1. Each P2P link will consume a /31 subnet and each loopback will consume a single address. Assuming that there will be 1 link between your borders, two uplinks per intermediate and two uplinks per switch stack that will consume 75 addresses. If you want to use the same pool for border handoff this will also consume some addresses(/30 per VN per border handoff). LAN automation will also not be able to start unless there is a minimum of 64 available addresses in the pool. I would probably reserve a /24 pool for LAN automation for this site. If you have plenty of addresses or expect expansion of the fabric you should reserve a /23. I recommend reading through the IP Pool planning portion of this doc: https://www.cisco.com/c/en/us/td/docs/cloud-systems-management/network-automation-and-management/dna-center/tech_notes/b_dnac_sda_lan_automation_deployment.html#id_89821 

2. No, they can be the same. If you are running 2.3.7.X you can manually set loopback addresses, if not it will be selected automatically from your main/principal IP pool.

3 & 4 & 5. There is no need for this, unless you wish to have separate ip ranges for troubleshooting purposes.

[thenetadmin]

Hi @Torbjørn 

Thanks.

If I may ask what would be your approach if you have similar design or setup?

[Torbjørn]

I would do the following assuming that your deployment only consists of this fabric and have plenty of available RFC1918 address space:

• Create a /22 global pool named "Underlay"
• Reserve a /23 pool for LAN automation
• Reserve a /24  pool for border handoff automation
• Set aside a /24 for management addresses(to be manually assigned during LAN automation)

The simplest approach would be to just use 1 pool for all of this, but by separating it like this makes troubleshooting a bit simpler.