用于连接性能改进的 Spark-SQL 查询提示

Question

I have recently been introduced to SparkSQL . We use Spark 2.4 . I recently found out that SparkSQL query supports the following hints for its Join strategies:

• BROADCAST hint
• MERGE hint
• SHUFFLE_HASH hint

Unfortunately, I have not found any online materials which elaborately discuss these hints and their application scenarios. I wish to learn some tips regarding when to use these hints in a query Join for improving query performance.

Can anyone explain with some examples. Any help is appreciated. Thanks

Answer 1

1. Broadcast join is very high performance join with sending data of the small table to every executor to execute a map-side join. here is the configuration:spark.sql.autoBroadcastJoinThreshold
2. Sort-merge join is a default join choice after spark 2.3

there are some post,Hope it help you: Spark SQL Joins Sort-Merge Join