Maybe I'm wrong here, but limiting data collection to a single NoSQL DB instance seems a little counter intuitive, especially in environments that have multiple polling engines. The idea is to allow muliple NoSQL DBs for NTA traffic, all of which roll up to a central DB, so that I can put NoSQL DBs closer to the additional pollers doing the NTA traffic collection. This would be especially useful over a high latency link or over a link that has lower bandwidth available between it and the central DB but a large number of Netflow sources on site.
This is tied to the idea I posted for NPM about Store and Forward Data Collection. In my mind, the NoSQL option should be made available for all modules with a central DB aggregation point.