SDA: Edge node behind Edge node

[markus.forrer]

Good morning 

Just a simple question about the physical design of a sd-access topology.

Is it supported that an Edge Node is attached to another Edge Node ?

Kind regards

[jedolphi · Accepted solution]

Yes, you can build a point-to-point routed link between FE1 and FE2, and have them in a chain, that is allowed. It is fine on 3650, 3850, 9300 and 9400 because on these platforms any switch port can be fabric facing. For 4500, having FEs in a chain is possibly a problem, since in 4500 we must have fabric facing ports on supervisor only, and it must be Supervisor 8-E or Supervisor 9-E. Best regards, Jerome

[ulasinski]

You can build it but what about the optimization of such a solution. Whether in campus fabric (LISP/Geo-VXLAN) should not be built according to the rules appropriate for SPINE-LEAF EVPN. I mean the same length of routes/hops (between each two itr/etr) and predictable network behavior?

[jedolphi]

Good day ulasinski,

SDA solves campus LAN use cases. In campus LAN it's not always possible to have symmetrical physical network topology, the realities of large physical campus with optical paths between disparate areas means we might have multiple tiers of network infrastructure, asymmetry, and sometimes SDA fabric edges in chains. That said, I would not tell someone to go out of their way to try to build a daisy chain of FEs. If possible then build symmetry. If not possible then build what's necessary. Make sure there is sufficient bandwidth in underlay and let the underlay IGP take care of ECMP and convergence between FEs. In SDA architecture a chain of FEs works and it is supported. Symmetry may be more optimal, sure, but it won't be achievable in all real world campus network e.g. an airport where running kilometers of optical fibre under busy roads / runways is not an option.

Best regards, Jerome

[imad-belal]

Dear  Jerome

Would you please, confirm if the Cat 9500 support the chain FE.

( Cat 9300 as fabric Edge -to- Cat 9500 Fabric Edge) 

Best 

Imad 

[jedolphi]

Hello Imad,

Please refer to compatibility matrix for 9500 models and code versions supported as FE,

https://www.cisco.com/c/en/us/solutions/enterprise-networks/software-defined-access/compatibility-matrix.html

If the 9500 is supported as FE, then it will work as FE in daisy chain.

Regards, Jerome

[Kevm]

Hi Jerome,

1. Would this design support the second layer fabric edge node uplink to two different Fabric Edge node for redundancy purposes?

Any restrictions/caveat on this design for Catalyst 9200 series?

2. How would the RLOC to EID map look like?

Thanks and regards

Kevin

[jedolphi]

Hi Kevin,

1. As long as we have fabric node Loopback to Loopback connectivity in SDA underlay (global routing table), then no issue. So this is fine. Use only p2p routed links between FEs please.

2. 9200 is an Fabric Edge switch, so it will have it's own specific RLOC (Lo0), same as every other FE.

Hope that makes sense, Jerome

[Kevm]

Thank you for the reply Jerome. Could I trouble you to point me to any documentation on supported FEs for daisy chain, caveats and design guide doc?

It's not available in https://www.cisco.com/c/dam/en/us/td/docs/solutions/CVD/Campus/sda-sdg-2019oct.pdf

Regards

Kevin

[jedolphi]

Hi Kevin, as far as I know there's no document to explain this because it's nothing exceptional or different. It's just loopback to loopback (FE to border, or FE to CP, or FE to FE) connectivity from a signalling and VXLAN perspective - I cover this tangentially and briefly in BRKCRS-3493 ( www.ciscolive.com ) where we talk about fabric over brownfields network - it's more or less same thing - routing between fabric nodes that doesn't care about the number of hops in the path between those fabric nodes  (there is other things to care about like MTU or RTT or multicast, but that has nothing to do with qty of hops in the path). From a limitations / caveats perspective, I'm not aware of any other than the obvious e.g. daisy chain of FEs might put more traffic on a given single uplink e.g. in scenario BDR--FE1--FE2--FE3--FE4, the BDR-FE1 link is carrying traffic of FE2,3 and 4 to border, but that's just common sense, and we can add another uplink and do ECMP or increase the speed of the existing uplink. If you're particularly concerned about caveats or gotchas you could start a conversation with your presales team, but my guess is they'll probably tell you the same, cheers, Jerome

[Kevm]

Thanks! I ll check on BRKCRS 3493.

[techno.it]

@jedolphi 

Do we have to configure this P2P links manually or it can be automated by DNAC?

[jedolphi]

Hello techno.it , if the fabric has been built on LAN Automation then you can automate the new link with LAN Automation, otherwise configure it manually. I haven't tested a DNAC template myself, but I believe that should work also. Cheers, Jerome