vRNI doesn't have any kind of native HA functionality so native vSphere HA or other external redundancy mechanisms are what you'd use to protect it.
The clustering option for platform VMs is only there to scale out and still designates one of them as primary for the UI so if that one goes down you'll be unable to access vRNI as described in the Clustering - General section of the FAQ guide.
Regarding #2, what you've pictured is one of the more common deployment models. The need for multiple proxies is technically driven more by scale than proximity since they're not really pushing a ton of traffic between DCs either way, so if you're under 100k flows per second you could also just run everything through a single proxy.
Just curious to know the use-case or requirement for vRNI - HA design
my understanding is - this one is analyzer tool
i have additional questions that
1. how many kinds of data source does vRNI support?
i only see vCenter, NSX, firewall and converged infrastructure from the data source drop down list.
(does that also mean the vRNI can only analyze NSX traffic?)
2. how do platform and proxy connect remotely?(e.g. platform at US site and proxy at Europe site)
does anyone know?
thanks in advance.
vRNI supports Physical as well as virtual data sources, please refer to docs for supported physical devices.
vRNI can analyze other traffic outside of NSX as well as it can support DVS,NSX, and physical flows
Platform and Proxy are linked during deployment please refer to documentation
Please refer to this guide for clustering
WIth the large amount of data that is passed to the platform it would be something to consider having the collector in another data center, but if your links can handle it, then it would be something to think about
is there any impact to business during vRNI operation specifically when enabling Netflow like traffic congestion or latency?
is there any overload on vDS processing?
No, there's no impact to production traffic when enabling netflow for vRNI. Netflow data generates a minimal amount of traffic is actually sent from the management vmkernel port so doesn't touch the logical interface that VMs in the associated port groups use.