I’m looking at building a customer data platform on top of AWS that can support both analytical and real-time use cases.

At a high level, events can flow into storage as well as trigger actions:

                         analytical stuff
                               |^
                               v|
       |---> processing ---> storage 
       |                       | ^
events |                       | |
       |                       v |
       |---> processing -> enrichment -> actions (200ms)

For example, segmenting (non real-time) the data or enriching incoming (real-time, ~200 ms).

Redshift (OLAP) appears to suit the analytical side, however does not suit the real-time aspect of data enrichment. RDS on the other hand seems to handle the real-time stuff just fine (OLTP) but not the analytical.

For the storage component of the diagram, would it be better to go with Redshift + some in-memory cache? Or is there a catchall for this?