Our install looks something like this:
the main app server is running all of the netflow, web, and NCM processes; the additional pollers do almost all of the UDT and NPM polling
I'm evaluating moving from 3.11 to 4.1x and scaling our install to deal with about 150,000 flows per second.
a) can I move the netflow listener to a dedicated box or can i run additional netflow received on the pollers to boost the receiving throughput?
b) that works out at about 1.2TB of storage for the flow data; does the product do sharding across multiple disks, or do I need to do that at the controller level?
c) do you think I'm crazy to try this?
if you've a large NTA install I;'d be interested in your experiences.
Those are great questions Richard... I don't have that many flow but... I can tell you having the flow database on the new noSQL database server is awesome! I now have a full month of up the the minute flow data which is huge improvement over the old system in the SQL Server. I've never had an issue with overloading the system yet. My servers are all virtualized now as well including the databases SQLServer and the noSQL flow database.