Storage

This is the stable storage to which intermittent or end results and alerts are written into. It's a very crucial component in the NRT context because we need to store the end results in a persistent store. Secondly, it serves as an integration point for further downstream applications which draw data from these low latency stores and evolve further insights or deep learning around them.

The following table clearly captures the various data stores and their alignment to the time SLA of NRT applications:

I would like to add a note here that we are skipping the plethora of options available for storage and visualization for now, but will touch upon these specifically in later sections of the book.