Out of the box, Cassandra is used internally for modeling data only and event data. Usage of Cassandra for customer data is possible, but not all use cases support the Cassandra data source. The most notable is that segmentation is only supported through customer data existing in a relational DB. If your use cases are entirely realtime/inbound, then pure Cassandra data source is a possibility. If you expect to use Segmentation as part of the solution, then you will need a data replica in both RDBMS and Cassandra. This is often useful to ensure a separation of batch intensive processes (in RDBMS) and realtime (Cassandra). For this configuration, RDBMS is typically the primary system of record and Cassandra is fed customer updates via data-flows.
I have a question for you Saleem_A. Can we store the data into Cassandra for every customer decision and decide to read the data back in response flow. Does this use case supported in delayed learning? Please let me know.
Reason for storing the data into Cassandra stores:We have more than 500 predictors in our application and we cannot expose the Pega marketing customer class with all the columns. The other option is to use a database view but it is not recommended by our application DBA.