Use the following recommendations to improve Spark performance with cloud data:
Accelerating ORC and Parquet Reads
Accelerating Sequential Reads Through Files in S3