Which nodes are required to communicate with external data sources before and after Fusioninsight HD Loader job?

Hive supports normal views and materialized views.

The role of Channel is similar to a queue, which is used to temporarily buffer incoming data. When Sink successfully sends the data to the next-hop Channel or the final destination, the data will be removed from the Channel.

Spark's intermediate data is stored in memory, which is more efficient for iterative calculations and batch calculations, and has higher latency.

Which of the following descriptions about Fusioninsight CTBase is wrong?