How to scale the database at runtime / Make the database available as a service - Over CF

Question

Let me set the premise . I am creating an API development platform that should be able to scale at runtime - on demand.

I am trying to improve the design of my system or trying to identify a possible performance bottleneck that might hit.

For the platform, I use the Cloud Foundry-based stack, which basically provides me with the ability to connect and play DB as a service using service connectors. Now, assuming that CF is doing it well, it seems so far, what should I do when the load suddenly increases?

Parameters:

• Create a fairly large cluster and list all services on it using CF DB as a service through service connectors (or the UPSI layer).
• Create a separate database as a platform to which CF will become a subscriber.
• Other methods are welcome! You might want to add the right design strategy as an answer.

I know how to implement the 1st point. I am interested in using Aerospike and Cassandra.

I want to study points 2 and 3. Recommendations are welcome!