Hi Authors,
In section 4.2 Tenant Network with Multiple Subnets
>From looking at deployment it looks like customer is just deploying L2 vxlan
>on the Leafs and Spines or super spine are connecting multiple Datacenter
>using L3 with East west traffic.
Can we also add other deployment scenario(s) which are deployed at scale.
Scenario1:
In reality we are seeing for Inter-subnet traffic is going to Super spine and
getting routed and down the spine and reaching the right host attached to Leafs.
I am seeing deployment like below fo
SS1 SS2 (super spine) (NVE Endpoints) -> multiple super spines
/ | \\/ \\
/ | /\\ \\ (ECMP paths)
/ |/ \\ \\
S1 S2 S3 (spine) -> multiple spines
|| || || (Underlay ECMP)
|| || ||
L1 L2 L3 (Nve Endpoints) - Leafs ->
| |
H1 H2
Figure 1 Example Network
Intra subnet traffic are bridged but they also reach Super spine if Spines are
connected in different pods.
Inter-subnet traffic are always routed to Super spine and they are forwarded by
L3 routing to correct spine within or across the PODs.
--
2nd scenario that's deployed at scale are
SS1 SS2 (super spine) -> multiple super spines
/ | \\/ \\
/ | /\\ \\ (ECMP paths)
/ |/ \\ \\
S1 S2 S3 (spine) -> multiple spines
|| || || (Underlay ECMP)
|| || ||
L1 L2 L3 (Nve Endpoints) - Leafs -> L2/L3 Vxlan deployed
(This allow number of routes to be distributed throughout leafs).
| |
H1 H2
In This scenario VIRB is in Leafs itselfs.
3rd. Can we also introduce controller in one of the use case scenario.
Thanks,
Deepak
_______________________________________________
nvo3 mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/nvo3
