Using Spark to save data to Cassandra

Question

Now in my current architecture I have a module which is responsible for writing/reading data to and from Cassandra, and module responsible for downloading data. Recently I started using Datastax and Spark. I want to do some transformations on new acquired data. What's the right take on this problem? Do I use my module for storing data and do Spark calculations separately, or send downloaded data directly to Spark using Spark Streaming and in a job save both the orginal data and transformed data to Cassandra? I'm operating on stock quotes so it's a lot of data downloaded continuously and a lot of transformations.

Answer 1

In my opinion, its better to keep it separated.

first store the raw data then process it.
its easier to scale and maintain each component later.

for example: if you want to change something in your downloading module like adding a new download sources or fix a bug, it wont affect the data processing done in spark, and changing something in the code running on spark wont have any effect(or introduce a bug) on the raw data you downloaded.