SDA L2 Flooding Limitations

[Volker Fries]

Hi together,

has anyone an idea about the limitations with L2 flooding ?

Our customer has a separate infrastructure with lots of L2 broacast domains (~100). This construct ist used for various sensors, terminals, ip-cams, etc...

and some devices which may not speak IP. So my question is, does any paper or recommendations/limitations exist for using the L2 Flooding Feature in SDA with DNA-C.

[Scott Hodgdon]

Volker,

There is a section for L2 Flood at https://www.ciscolive.com/c/dam/r/ciscolive/apjc/docs/2020/pdf/BRKCRS-3493.pdf .

Ideally, we want all DHCP-capable devices to move to new subnets if possible and then route between SDA and non-SDA during migration. However, that is is not always possible, so we support L2 Selective Flooding ("selective" because it is on a per IP Pool basis and selected when an IP Pool is added to a VN) and L2 Handoff on a border. L2 Selective Flooding requires multicast to be enabled in the underlay , and this can be done either by LAN Automation or manually.

There are a few important considerations with L2 handoffs:

1. An L2 handoff for a subnet can exist on only one Border at a time. For resiliency, most customers would use a StackWise Virtual pair of Catalyst 9500 High Performance switches, a chassis based system with multiple supervisors or a stack of Catalyst 9300s ( if scale is OK ) to host an L2 handoff. 

2. The scale of the Border hosting the L2 handoff differs by platform, and so the number of clients needing the L2 service both inside and outside the fabric will need to be known. If a single Border is not sufficient, additional Borders can be added to host L2 handoffs.

Cheers,
Scott Hodgdon

Senior Technical Marketing Engineer

Enterprise Networking and Cloud Group

[Volker Fries]

Hi Scott,

thanks for the information, so far.

But is there a practical limitation of clients within the L2 Flooding VN and is there a practicle number of max. VN´s to be useed for that ?

In our case we speak about ~ actual 100 L2 domains.

[Stevenmns]

Did you ever find an answer to this?

[Scott Hodgdon]

@Volker Fries and @Stevenmns ,

You can find L2 Handoff client scale limitations in Table 7 at https://www.cisco.com/c/en/us/products/collateral/cloud-systems-management/dna-center/nb-06-dna-center-data-sheet-cte-en.html . You can find Fabric VN Scale in Table 16. Remember that we can cary multiple IP Pools in one VN, so you should think about scale in that way as well. We don't necessarily need a VN for every VLAN that will be on an L2 Handoff.

Cheers,
Scott Hodgdon

Senior Technical Marketing Engineer

Enterprise Networking and Cloud Group

[moamen.elhefnawy]

Hello Scott,

I'm interested to find the cisco live session you mentioned above, can you share with me if you have it, as the link you shared is now broken, and the session not available now on Cisco live.

https://www.ciscolive.com/c/dam/r/ciscolive/apjc/docs/2020/pdf/BRKCRS-3493.pdf

Thanks & BR

Moamen 

[balaji.bandi]

Cisco may have removed the on-demand older presentation check below that helps you :

https://www.ciscolive.com/c/dam/r/ciscolive/emea/docs/2025/pdf/BRKENS-2824.pdf

[moamen.elhefnawy]

Hello,

Thanks for your reply, actually I'm interested in this exact session as it should include some info about how to cascade multiple Fabric edge nodes behind each other, but I can't find this info in the session you shared.

But thanks anyway , still usueful one.

Thanks & BR

Moamen 

[Andrii Oliinyk]

as soon as u have 1st extndNode connected u can cascade downstream extended node connected in the same manner with certain depth limit

[balaji.bandi]

how to cascade multiple Fabric edge nodes behind each other, but I can't find this info in the session you shared.

as long as the extended switch is fabric supported you can extend daisy chained 

Example : CORE-FBS-FIS-FES-FEN(this is extended node)

[Andrii Oliinyk]

cascading ENs is as easy as LAN-A. u simply select 2 ENs as seeds & voi-la

[Andrii Oliinyk]

XENs & PENs play by STP rules. with depth of cascade u are limited by STP diameter. hence in some scenarios u'd prefer to replace stp chain with rpr or mrp (not sure what catc can do on this)