How much should one choose the number of executors and cores on a spark-submit job?

Question

I have a spark structured streaming job that does the following:

1. Streams from S3 Folder a file containing json (many json